Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Title Study on Upward Turbulent Bubbly Flow in Ducts(Dissertation_全文 )
Author(s) Zhang, Hongna
Citation Kyoto University (京都大学)
Issue Date 2014-09-24
URL https://doi.org/10.14989/doctor.k18590
Right 許諾条件により要旨は2014/12/24に公開
Type Thesis or Dissertation
Textversion ETD
Kyoto University
Study on Upward Turbulent Bubbly Flow in Ducts
ZHANG HONGNA
Doctoral Dissertation
Department of Nuclear Engineering
Graduate School of Engineering
Kyoto University
September 2014
i
Acknowledgements
I here would like to sincerely thank many people for their kind aids and
encouragements during my doctor course in Kyoto University.
Foremost, I would like to express my special appreciation and thanks to my
supervisor Profess Tomaki KUNUGI, who has been a tremendous mentor of me for his
patience, motivation, enthusiasm, and immense knowledge. I would like to thank him
for showing me what the research is and how to enjoy it, for discussing on many
scientific problems, for encouraging me, and for allowing me to grow as a research
scientist freely. I have been extremely lucky to have a supervisor like him.
Special thanks are also extended to my committee members, Professor
Kazuyoshi NAKABE and Associate professor Takehiko YOKOMINE for serving as
my committee members and for their questions and suggestions during my defense.
I also wish to thank all the members in KUNUGI Lab for their help on my life
and research in Kyoto University. Special thanks are given to Professor Jun
SUGIMOTO, Lecturer Zensaku KAWARA, Dr. Li-Fang JIAO, Dr. Hao-Min SUN, Dr.
Hong-Son PHAM and Dr. Yang CAO for their help on my life and their discussions on
my research and Ms. Hijiri SOEJIMA for her management on my financial budget.
Thanks are also extended to Japan Society for the Promotion of Science (JSPS)
Research Fellowship and Global Center of Excellence (G-COE) Program (J-051) for
their financial support during my doctor course.
Finally, special thanks are given to my family and my friends. Words cannot
express how grateful I am to my father Mr. Liu-Xian ZHANG, my mother Mrs.
Feng-Ying LI, and my brother Mr. Chao ZHANG for all their sacrifices,
encouragements and supports that they’ve made on my behalf. I would also like to
thank all of my friends who supported me in writing, and encouraged me to strive
towards my goal.
ii
iii
Table of Contents
Acknowledgements .......................................................................................................... i
List of Tables ................................................................................................................. vii
List of Figures .............................................................................................................. viii
Chapter 1 Introduction ............................................................................................... 1
1.1 Motivation .......................................................................................................... 1
1.2 State-of-the-art of turbulent bubbly flow............................................................ 2
1.2.1 Background .................................................................................................. 2
1.2.2 Review of the experimental and numerical databases ................................. 5
1.2.3 Review of existing models for turbulent bubbly flow ............................... 17
1.2.4 Summary on review of databases and models ........................................... 28
1.3 Scope and layout of the dissertation ................................................................. 29
Reference .................................................................................................................... 30
Chapter 2 Experimental databases of upward turbulent bubbly flow ................. 37
2.1 Introduction ...................................................................................................... 37
2.2 Experimental techniques................................................................................... 37
2.3 Experimental apparatus .................................................................................... 38
2.4 Experimental results ......................................................................................... 42
2.5 Summary ........................................................................................................... 43
Reference .................................................................................................................... 51
Chapter 3 Numerical evaluation of two-fluid model of turbulent bubbly flow in
large square duct .......................................................................................................... 53
3.1 Introduction ...................................................................................................... 53
3.2 Numerical methodology ................................................................................... 53
3.2.1 Problem description ................................................................................... 53
3.2.2 Numerical methods .................................................................................... 54
iv
3.3 Results and discussion ...................................................................................... 59
3.4 Summary ........................................................................................................... 60
Reference .................................................................................................................... 63
Chapter 4 Developing process of turbulent bubbly flow ....................................... 65
4.1 Introduction ...................................................................................................... 65
4.2 Experimental observations ............................................................................... 65
4.3 Modeling of developing process ...................................................................... 71
4.3.1 Formation of bubble layer in the first stage ............................................... 71
4.3.2 Liquid phase velocity lifting in the second stage ...................................... 74
4.4 Summary and Discussion ................................................................................. 77
Reference .................................................................................................................... 78
Chapter 5 Axial liquid-phase velocity of turbulent bubbly flow ........................... 79
5.1 Introduction ...................................................................................................... 79
5.2 Index of flow type .............................................................................................. 80
5.3 Conditions for M-shaped Wl profile ................................................................... 89
5.4 Duct geometry effect ........................................................................................ 90
5.5 Peak location of Wl near the wall ..................................................................... 95
5.6 Flow characteristics in two layers .................................................................... 95
5.6.1 Definition of two layers ............................................................................. 96
5.6.2 Velocity characteristics .............................................................................. 97
5.7 Numerical validation of the liquid phase velocity .......................................... 101
5.7.1 Numerical models .................................................................................... 101
5.7.2 Numerical results ..................................................................................... 103
5.8 Summary and discussion ................................................................................ 105
Reference .................................................................................................................. 107
Chapter 6 Void fraction of the turbulent bubbly flow ......................................... 109
v
6.1 Introduction .................................................................................................... 109
6.2 Bubble deformation ......................................................................................... 110
6.2.1 Definition of bubbles’ statistical shape ..................................................... 110
6.2.2 Distribution of bubbles’ shape factor ........................................................ 117
6.2.3 Void fraction effects on shape factor ....................................................... 121
6.2.4 Bubble size validation ............................................................................. 127
6.3 Numerical validation of the void fraction....................................................... 129
6.3.1 Numerical models .................................................................................... 129
6.3.2 Numerical results ..................................................................................... 131
6.4 Summary and Discussion ............................................................................... 136
Reference .................................................................................................................. 137
Chapter 7 Turbulence modulation of the turbulent bubbly flow ........................ 139
7.1 Introduction .................................................................................................... 139
7.2 Turbulent kinetic energy ................................................................................. 140
7.2.1 Bubble-induced turbulence at the duct center ......................................... 147
7.2.2 Bubble-induced and wall-induced turbulence interaction ....................... 154
7.3 Turbulence anisotropy in turbulent bubbly flow ............................................ 166
7.3.1 Bubble effects on turbulence anisotropy in liquid phase ......................... 166
7.3.2 Modeling of the anisotropic turbulent bubbly flow ................................. 174
7.4 Summary and discussion ................................................................................ 180
Reference .................................................................................................................. 182
Chapter 8 Conclusions ............................................................................................ 185
8.1 Summary of dissertation ................................................................................. 185
8.2 Recommendation for future work................................................................... 187
Nomenclature .............................................................................................................. 189
vi
vii
List of Tables
Table 1.1 Numerical databases based on DNSs of the turbulent bubbly flow.
Table 2.1 Experimental databases in different duct geometries. B: Base; H: Height.
Table 2.2 Experimental flow conditions in different duct geometries. B: Bubbly flow;
T: Transition regime.
Table 4.1 Experimental conditions for the developing process.
Table 5.1 Experimental data of dp/dz (kPa/m).
Table 5.2 Geometry dimension H to determine the wall shear stress
Table 5.3 Experimental conditions to for different geometries in the literatures. In the
table, C: circular duct; T: triangle duct; S: square duct; B: base length; H:
height.
viii
List of Figures
Fig. 2.1 (a) Experimental techniques for the flow measurements and (b) its
applications in the turbulent bubbly flow.
Fig. 2.2 Simplified schematic diagram of the experimental apparatus.
Fig. 2.3 Lateral distributions of axial liquid phase velocities Wl and void fraction α
in the small circular pipe with DH=60mm under (a) Jl =0.74m/s and (b) Jl
=1.03m/s based on Serizawa’s experimental database [2.1].
Fig. 2.4 Lateral distributions of axial turbulent intensities <ww> in the small circular
pipe with DH=60mm based on Serizawa’s experimental database: under (a)
Jl =0.74m/s and (b) Jl =1.03m/s [2.1].
Fig. 2.5 Lateral distributions of axial liquid phase velocities Wl and void fraction α
in the small circular pipe with DH=50.8mm based on Hibiki’s experimental
database: under (a) Jl =0.491m/s and (b) Jl =0.986m/s [2.6].
Fig. 2.6 Lateral distributions of axial turbulent intensities <ww> in the small circular
pipe with DH=50.8mm based on Hibiki’s experimental database: under (a)
Jl =0.491m/s and (b) Jl =0.986m/s [2.6].
Fig. 2.7 Lateral distributions of axial turbulent kinetic energy <ww> and lateral
turbulent kinetic energy <uu> in the small circular pipe in the small circular
pipe with DH=60mm based on Liu’s experimental database [2.3-2.4].
Fig. 2.8 Lateral distributions of axial liquid phase velocities Wl and void fraction α
in the large circular pipe with DH=200mm based on Shakwat’s experimental
database: under (a) Jl =0.45m/s and (b) Jl =0.68m/s [2.7].
Fig. 2.9 Lateral distributions of axial turbulent kinetic energy <ww> and lateral
turbulent kinetic energy <uu> in the large circular pipe with DH=200mm
based on Shakwat’s experimental database: under (a) Jl =0.45m/s and (b) Jl
=0.68m/s [2.7].
Fig. 2.10 Lateral distributions of axial liquid phase velocities Wl and void fraction α
ix
in the large circular pipe with DH=136mm based on Sun’s experimental
database: under Jl =0.75m/s (a) along the bisector line; (b) along the
diagonal line; under Jl=1.00m/s (c) along the bisector line; (d) along the
diagonal line [2.10].
Fig. 2.11 Lateral distributions of axial turbulent kinetic energy <ww> and lateral
turbulent kinetic energy <uu> in the large circular pipe with DH=136mm
based on Sun’s experimental database: under Jl =0.75m/s (a) along the
bisector line; (b) along the diagonal line; under Jl=1.00m/s (c) along the
bisector line; (d) along the diagonal line [2.10].
Fig. 3.1 Schematic diagram of the numerical problem.
Fig. 3.2 Flow charts of numerical calculation.
Fig. 3.3 Comparisons of the experimental measurements and numericla predictions
under Jl =0.75m/s and Jg =0.09m/s: for the liquid phase velocity (a) along
the bisector line and (b) along diagonal line; for the void fraction (c) along
the bisector line and (d) along diagonal line; for turbulence (e) along the
bisector line and (f) along diagonal line;
Fig. 4.1 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5
DH, 10 DH and 16DH for Jl=0.5m/s: with Jg=0.09m/s along (a) the bisector
line; (b) the diagonal line; with Jg=0.135m/s along (c) the bisector line; (d)
the diagonal line [4.7].
Fig. 4.2 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5
DH, 10 DH and 16DH for Jl=0.75m/s: with Jg=0.09m/s along (a) the bisector
line; (b) the diagonal line; with Jg=0.135m/s along (c)the bisector line; (d)
the diagonal line; with Jg=0.18m/s along (e) the bisector line; (f) the
diagonal line [4.7].
Fig. 4.3 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5
DH, 10 DH and 16DH for Jl=1.0m/s: with Jg=0.09m/s along (a) the bisector
x
line; (b) the diagonal line; with Jg=0.135m/s along (c) the bisector line; (d)
the diagonal line; with Jg=0.18m/s along (e) the bisector line; (f) the
diagonal line [4.7].
Fig. 4.4 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5
DH, 10 DH and 16DH for Jl=1.25m/s: with Jg=0.135m/s along (a) the
bisector line; (b) the diagonal line; with Jg=0.18m/s along (c) the bisector
line; (d) the diagonal line; with Jg=0.225m/s along (e) the bisector line; (f)
the diagonal line [4.7].
Fig. 4.5 Schematic diagram of the developing process of the upward turbulent
bubbly flow in the large square duct.
Fig. 4.6 Schematic diagram of the formation of the bubble layer due to the lift force
exerted on the bubbles and the developing of the boundary layer in the
liquid phase.
Fig. 4.7 Comparison between the thickness of the boundary layer δl and the bubble
layer δB at the measurement location z=4.5DH.
Fig. 4.8 Schematic diagram of liquid phase velocity lifting up in the second stage.
Fig. 4.9 Comparison of dimensionless parameter 2/ 1pA L g W for
different superficial liquid and gas phase velocities Jl and Jg and
measurement locations.
Fig. 5.1 Schematic diagram of the type of the velocity profile of the upward flow:
(a) M-shape; (b) ∩-shape; (c) Cuspid-shape; (d) W-shape
Fig. 5.2 Distributions of the local net driving force F for (a) Jl=0.75m/s and (c)
Jl=1.0m/s and corresponding liquid velocities Wl for (b) Jl=0.75m/s and (d)
Jl=1.0 m/s along the bisector line.
Fig. 5.3 Distributions of the local net driving force F for (a) Jl=0.75m/s and (c)
Jl=1.0m/s and corresponding liquid velocities Wl for (b) Jl=0.75m/s and (d)
Jl=1.0m/s along the diagonal line.
xi
Fig. 5.4 Schematic diagram of flow image for the small void fraction.
Fig. 5.5 Schematic diagram of flow image for the large void fraction.
Fig. 5.6 Schematic diagram of velocity peak location.
Fig. 5.7 Comparison of γ for different duct geometries for Jl around (a) 0.75m/s and
(b) 1.0m/s. For noncircular geometry, ―-1‖ and ―-2‖ indicate data obtained
near the wall center and corner, respectively.
Fig. 5.8 Definition of two layers
Fig. 5.9 Velocity characteristics in the inner layer along (a) the bisector line and (b)
diagonal line; and in the outer layer (c) along the bisector line and (d) the
diagonal line. The mean void fraction could be approximated as
αgm=Jg/(Jg+Jl).
Fig. 5.10 Estimation of lB-B from the experiments. T, TB and TL are total measuring
time, bubbles passing time and liquid passing time, respetively. WB is the
mean bubble velocity.
Fig. 5.11 Comparison between numerical predicted liquid phase velocity Wl and
experimental measurement on the bisector line for (a) Jl =0.75m/s and (b) Jl
=1.0m/s, respectively.
Fig. 5.12 Lateral distribution of mean viscous force ηv based on the experimental
measurements for (a) Jl =0.75m/s and (b) Jl =1.0m/s, respectively.
Fig. 6.1 The distributions of the Sauter mean diameter DSM and the chord length Lc
along the bisector line under different flow conditions :(a) Jl = 0.5m/s; (b) Jl
= 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s. The solid and dashed lines show
DSM, and Lc respectively.
Fig. 6.2 The distributions of the Sauter mean diameter DSM and the chord length Lc
along the diagonal line under different flow conditions :(a) Jl = 0.5m/s; (b)
Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s. The solid and dashed lines
show DSM, and Lc respectively.
xii
Fig. 6.3 The distributions of the Sauter mean diameter DSM and the chord length Lc
along the near-wall line under different flow conditions :(a) Jl = 0.5m/s; (b)
Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s. The solid and dashed lines
show DSM, and Lc respectively.
Fig. 6.4 Schematic diagram of the ellipsoidal bubble shape.
Fig. 6.5 The aspect ratio E vs. the ratio DSM/Lc for the ellipsoidal bubbles.
Fig. 6.6 The distributions of ratio r along bisector line under different flow
conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s.
Fig. 6.7 The distributions of the ratio r along the diagonal line under different flow
conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s.
Fig. 6.8 The distributions of the ratio r along the near-wall line under different flow
conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl = 1.25m/s.
Fig. 6.9 The distributions of the ratio E/E0 along the bisector line under different
flow conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl =
1.25m/s.
Fig. 6.10 The distributions of the ratio E/E0 along the diagonal line under different
flow conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl =
1.25m/s.
Fig. 6.11 The distributions of the ratio E/E0 along the near-wall line under different
flow conditions: (a) Jl = 0.5m/s; (b) Jl = 0.75m/s; (c) Jl = 1.0m/s; (d) Jl =
1.25m/s.
Fig. 6.12 The ratio R=Eα/ES vs. the liquid fraction (1-α) under all the tested flow
conditions.
Fig. 6.13 DH.exp vs. DHS under all the tested flow conditions.
Fig. 6.14 DH.exp vs. DHα under all the tested flow conditions.
Fig. 6.15 Schematic diagram of the turbulent components along the diagonal line.
Fig. 6.16 Comparison between numerical predicted void fraction and experimental
xiii
measurement for Jl =0.5m/s at z=16DH.
Fig. 6.17 Comparison between numerical predicted void fraction and experimental
measurement for Jl =0.75m/s at z=16DH.
Fig. 6.18 Comparison between numerical predicted void fraction and experimental
measurement for Jl =1.0m/s at z=16DH.
Fig. 7.1 Lateral distributions of the ratio of the turbulent kinetic energies between tkt
and tks in the small circular pipe with DH=60mm based on Serizawa’s
experimental database: under (a) Jl =0.74m/s and (b) Jl =1.03m/s [7.2].
Fig. 7.2 Lateral distributions of the ratio of the turbulent kinetic energies between tkt
and tks in the small circular pipe with DH=50.8mm based on Hibiki’s
experimental database: under (a) Jl =0.491m/s and (b) Jl =0.986m/s [7.15].
Fig. 7.3 Lateral distributions of the ratio of the turbulent kinetic energies between tkt
and tks in the small circular pipe with DH=60mm based on Liu’s
experimental database [7.7].
Fig. 7.4 Lateral distributions of the ratio of the turbulent kinetic energies between tkt
and tks in the large circular pipe with DH=200mm based on Shakwat’s
experimental database: under (a) Jl =0.45m/s and (b) Jl =0.68m/s [7.13].
Fig. 7.5 Lateral distributions of the ratio of the turbulent kinetic energies between tkt
and tks in the large circular pipe with DH=136mm based on Sun’s
experimental database: under (a) Jl =0.75m/s along the bisector line; (b) Jl
=0.75m/s along the diagonal line; (c) Jl =1.0m/s along the bisector line; (b)
Jl =1.0m/s along the diagonal line [7.16].
Fig. 7.6 (a) The relation between the ratio tkt / tks and the local void fraction α in the
duct center for different ducts; (b) The critical flow conditions Jl and Jg for
tkt / tks ≈ 25.
Fig. 7.7 (a) The axial turbulent intensities <ww> vs. the local void fraction α in the
duct center; (b) <ww>/Wl vs. the local void fraction α in the duct center.
xiv
Fig. 7.8 Comparison the experimental data of the turbulent kinetic energy tk of the
liquid phase with the new correlation Eq. (7.7) for different bubble diameter
at the duct center.
Fig. 7.9 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to
that in the duct center c
ktt
in the small circular pipe with DH=60mm based
on Serizawa’s experimental database: under (a) Jl =0.74m/s and (b) Jl
=1.03m/s based on Serizawa’s experimental database [7.2].
Fig. 7.10 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to
that in the duct center c
ktt
in the small circular pipe with DH=50.8mm
based on Hibiki’s experimental database: under (a) Jl =0.491m/s and (b) Jl
=0.986m/s [7.15].
Fig. 7.11 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to
that in the duct center c
ktt
in the small circular pipe with DH=60mm based
on Liu’s experimental database [7.7].
Fig. 7.12 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to
that in the duct center
c
ktt in the large circular pipe with DH=200mm based
on Shakwat’s experimental database: under (a) Jl =0.45m/s and (b) Jl
=0.68m/s [7.13].
Fig. 7.13 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to
that in the duct center
c
ktt in the large noncircular pipe with DH=136mm
based on Sun’s experimental database: under Jl =0.75m/s (a) along the
bisector line; (b) along the diagonal line; under Jl =1.0m/s (c) along the
bisector line; (b) along the diagonal line [7.16].
Fig. 7.14 Comparison between , , ,/B w B w B c
kw kb kbt t t and /w c based on Sun’s
experimental database [7.16].
Fig. 7.15 (a) Relation between /w c and c ; (b) Relation between
, , ,/B w B w B c
kw kb kbt t t and c based on Sun’s experimental database [7.16].
xv
Fig. 7.16 Schematic diagram of the relation between the turbulence in the liquid
phase and the void fraction for the turbulent bubbly flow.
Fig. 7.17 Experimental conditions from [7.8] with turbulence suppression and
enhancement in the liquid phase.
Fig. 7.18 Comparison the experimental measured void fraction and the calculated
void fraction based on 2, /2 R
wS
kwcalc Wt near the wall.
Fig. 7.19 Schematic diagram of the mechanism for the turbulence suppression
Fig. 7.20 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=60mm based on Serizawa’s experimental
database: under (a) Jl =0.45m/s and (b) Jl =0.61m/s [7.2].
Fig. 7.21 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=57.2mm based on Liu’s experimental database
[7.7].
Fig. 7.22 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=57.15mm based on Wang’s experimental
database [7.6].
Fig. 7.23 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
large circular pipe with DH=200mm based on Shakwat’s experimental
database: under (a) Jl =0.45m/s and (b) Jl =0.68m/s [7.13].
Fig. 7.24 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
large circular pipe with DH=136mm based on Sun’s experimental database:
under (a) Jl =0.75m/s along the bisector line; (b) Jl =0.75m/s along the
diagonal line; (c) Jl =1.0m/s along the bisector line; (b) Jl =1.0m/s along the
diagonal line [7.16].
Fig. 7.25 The relation between the turbulence anisotropy tensor b11 and b33 and the
local void fraction α in the duct center for different ducts. The black and
green symbols indicate the results in the small circular pipes and the large
xvi
ducts respectively.
Fig. 7.26 The relation between the turbulence anisotropy tensor b11 and b33 and the
local bubble size α in the duct center for different ducts. The black and
green symbols indicate the results in the small circular pipes and the large
ducts respectively.
Fig. 7.27 (a) Contribution of each term in the energy balance to the turbulence
anisotropy; (b) Schematic diagram of the generation of the turbulence
anisotropy of the single-phase turbulent flow in channel.
Fig. 7.28 Schematic diagram of grid-generated decaying homogeneous isotropic
turbulence.
Fig. 7.29 Schematic diagram of turbulence generation of the turbulent bubbly flow in
the core region where the turbulence is dominated by bubbles.
1
Chapter 1
Introduction
1.1 Motivation
Nowadays, in a wide variety of engineering systems the upward turbulent
bubbly flows play an increasingly important role such as boiling water and pressurized
water nuclear reactors (BWRs and PWRs), heat exchangers, petroleum transportation
systems and so on. Therein, accurate predictions of the flow characteristics are
essentially required for the design, process optimization and safety control. For example,
in the various subchannels of a nuclear fuel rod bundle of BWRs, the predictions of the
flow especially the void lateral distribution must be accurately carried out in order to
analyze the performance of the heat transfer such as local critical heat flux (CHF) which
is an important thermal-hydraulic phenomenon for the safety control. During the
transportation of the crude petroleum which is usually the two-phase or multiphase flow,
the predictions of the flow especially the drag friction are very crucial for the pipeline
design and operations such as the arrangement of the pump stations. Therefore, the
reliable two-phase flow models are in urgently needs based on the well understanding of
their physical processes.
So far, with the development of the experimental techniques and computational
fluid dynamics (CFD), numerous researches on the turbulent bubbly flow have been
carried out, on basis of which the understanding and modeling of turbulent bubbly flow
have been improved. However, the existence of the multi-deformable and moving
interfaces therein could induce the significant discontinuities of the fluid properties and
the complex flow field near the interface, which makes it much more complicated as
compared to the single-phase turbulent flow. Consequently, our current understanding
2
about the turbulent bubbly flows is still very limited due to its complicated properties,
and thus the predictive capability of the existing modeling is still far from the level of
the single-phase turbulent flow.
Due to its wide application in the various engineering systems and the lack of a
satisfactory knowledge about this complicated phenomenon, more researches related
with this phenomenon are necessary to be carried out in order to improve our physical
understanding and the existing models.
1.2 State-of-the-art of turbulent bubbly flow
In this section, the state-of-the-art of the upward turbulent bubbly flow will be
reviewed including the following topics:
(a) How were the previous researches on the turbulent bubbly flows carried out?
(b) What are the key issues in order to understand and model the turbulent
bubbly flow?
(c) What have we achieved based on the previous researches?
(d) What needs to be studied on the turbulent bubbly flow in the future?
1.2.1 Background
Generally, the researches to understand and model the turbulent bubbly flow
started from the single-phase turbulent flow which was considered as one of the most
complicated and unsolved physical problem in the last century and taken as the
background knowledge for the turbulent bubbly flow.
As for the single-phase turbulence, since the observation of the turbulence
phenomenon by Reynolds’s famous experiments in 1883, the innumerable researches
have been carried out from the experimental, theoretical and numerical point of views.
In general, the following topics were mainly focused on [1.1] including: (1) how the
turbulent flows behave; (2) what kinds of the fundamental physical processes involved;
(3) how to describe the turbulent flows quantitatively; (4) how to simulate and model
the turbulent flows to meet the requirements of the engineering application. Owing to
3
the contributors, our knowledge on single-phase turbulence has been improved
significantly in the past century.
As for the researches of the turbulent bubbly flows, the similar targets to that of
the single-phase turbulent flow are aimed at, which are solving the above four questions.
However, in the turbulent bubbly flows, there exist the multi-deformable and moving
interfaces which could also be coalesced to form new interfaces or broken into more
small interfaces [1.2]. Therefore, in addition to the general researches of the
single-phase turbulent flow, the bubbles behavior and bubbles’ effect are necessary to
be considered for the turbulent bubbly flow. Comparing to the single-phase turbulent
flow, the turbulent bubbly flow owns the different characteristics due to the existence of
the dispersed phase such as:
(a) Discontinuities of the fluid properties: During the mathematical description of
the turbulent bubbly flow, the general continuum fluid assumption used for
the single-phase flow becomes invalid because of the presence of the
multi-interfaces.
(b) Relative motion: Due to the density difference or the buoyancy effect, the
relative motions exist between the bubbles and the liquid phase.
(c) New characteristic scale: Additional characteristic scales related with the
bubble size DB are introduced on basis of the duct geometrical size DH and
eddy scales lv such as Kolmogorov scale and the integrated length scale le for
the single-phase turbulent flow.
(d) Phase interaction: Due to the relative motion and the new characteristic scale,
the flow characteristics in each phase could be influenced by the other phase
through the interfaces between them. For example, from the micro-scale point
of view, the presence of the interfaces could induce complex flow field near
the interface and consequently the complex flow field could modify the
interface characteristics such as bubble shape. For the turbulent bubbly flow
in the pipes, from the macro-scale point of view, the phase distribution could
be affected due the interaction between two phases. Consequently, the flow
4
field in the liquid phase such as velocities and turbulence distribution could
then be modified comparing to that of the single-phase flow. For a long time,
the understanding of the interfacial interaction, the phase distribution in the
turbulent bubbly flow and the turbulence modulation were very important
research topics and attracted numerous attentions.
(e) Complicated flow closure problem: With the interfacial interactions and
difficulties of the mathematical description, the flow closure problems of the
momentum and energy in the turbulent bubbly flow become more
complicated comparing to that of the single-phase turbulent flow. For
example, therein how to model the contributions of the phase interaction and
how to numerically solve the proposed models are both important research
topics.
Basically, for the turbulent bubbly flow, the existing researches from the
experimental, theoretical and numerical point of views were mainly focused on the
following questions aiming at accurate modeling of the turbulent bubbly flows.
(a) How does the turbulent bubbly flow behave? In addition to the general
researches of the single-phase turbulent flow, herein how does the bubbles
behave such as bubble distribution, deformation, coalescence and breakup
are also necessary to be considered.
(b) What kinds of the fundamental physical processes are involved in the
turbulent bubbly flow? In addition to the general researches of the
single-phase turbulent flow, herein what kind of the fundamental physical
processes responsible for the bubbles behaviors and the phase interactions
are also necessary to be considered.
(c) How to describe the turbulent bubbly flows quantitatively? In addition to the
general researches of the single-phase turbulent flow, herein how to solve the
problem induced by discontinuities of the fluid properties and how to
quantitatively describe the bubbles behaviors and phase interactions were
also necessary to be considered.
5
(d) How to simulate and model the turbulent bubbly flows to meet the
requirements of the engineering application? Additionally, how to model the
phase interaction based on the understanding the physical process are also
necessary to be considered.
In the following sections, what have we achieved about the above questions for
the turbulent bubbly flows will be reviewed based on the existing literatures. Firstly, the
establishment of the database of turbulent bubbly flow will be reviewed to answer the
first two questions mentioned above and validate the later proposed models. Then, the
development and implementation of the models for the turbulent bubbly flow will be
summarized to answer the latter two questions mentioned above. Finally, the unsolved
problems and the scope of the current dissertation will be given.
1.2.2 Review of the experimental and numerical databases
Currently, the databases for the turbulent bubbly flows are established based on
the experimental measurements and the direct numerical simulations (DNSs).
1.2.2.1 Experimental measurements
To understand the physical process and develop the model of the turbulent
bubbly flows, the detailed flow information such as the drag resistance, the temporal
and spatial evolutions of velocities and turbulence in two phases and the detailed
bubbles behaviors such as the bubble concentration, the bubble size, the bubble shape
and the bubbles motion are necessary at the same time. However, so far limited by the
current experimental techniques, for the turbulent bubbly flow, only several flow
parameters are available such as the statistical liquid phase velocities and turbulence
intensities, the void fraction, the bubble size and velocity, and the interfacial area
concentrations (IAC) and so on. The experimental databases on these parameters have
been established.
6
1.2.2.1.1 Database of turbulent bubbly flow
So far the databases of the turbulent bubbly flow have been established in the
following flow types:
(i) Uniform bubbly flow
The uniform flow is a simplest flow type for the understanding of the flow
characteristics and physical process and the developing the models. The studies on the
uniform turbulent flows such as the homogeneous turbulent flows generated by
homogeneous shear and the homogeneous isotropic turbulent flows generated by grids
have played an important role for the model development in the single-phase turbulent
flow [1.1]. As for the turbulent bubbly flow, so far Marie [1.3] and Lance et al.[1.4-1.5]
developed the measurements of the liquid-phase turbulence in a grid-generated and a
homogeneous shear-generated bubbly turbulent flow based on Laser-Doppler and
hot-film anemometry. Therein, the following liquid turbulent characteristics were
studied including the relation between the turbulent kinetic energy and the void fraction,
the isotropy level of the Reynolds stress tensor and the turbulent energy spectra.
According to their studies, two distinct regimes were found according to the void
fraction, which are a) for the low void fraction where hydrodynamic bubbles
interactions are negligible, the turbulent kinetic energy of the liquid could be considered
to be just the summation of the turbulent kinetic energy generated by the grid and the
motion of non-interactive bubbles; b) for the higher void fraction cases, due to the
bubbles mutual interactions, the bubbles transfer a greater amount of kinetic energy to
the liquid and the bubble-induced turbulence dominates the grid-generated turbulence.
In addition, during the above process, the isotropy of the turbulent field was found to be
not altered or improved by bubbles. The one-dimensional turbulent energy spectra of the
bubble-induced wake were found to be deviated from the classical power law exponents.
However, it is notable that in these experiments, the void fraction was very low less than
3% with the bubble diameter around 5mm. As for the other conditions, it still needs
further investigation.
7
(ii) Boundary layer flow
The boundary flow is another simple flow type for the understanding of the flow
characteristics and physical process and the developing the models for the single-phase
turbulent flow [1.1]. For the turbulent bubbly flow, so far there are very few researches
of this flow type, except Moursali et al.[1.6] developed the measurements in an upward
turbulent bubbly boundary layer along a vertical flat plate based on LDV. Therein, they
focused on the void fraction distribution, the wall shear stress, and the mean liquid
velocity profiles. According to their study, the lateral bubble migration toward to the
wall occurs depending on the bubble mean diameter and the void fraction similar to the
duct flow. In addition, the wall skin friction coefficient was observed to increase
because of the presence of the bubbles, which modifies the universal logarithmic law
near the wall. In this experiment, the measured location was mainly at 1m from the inlet
and the void fraction was less than 10% with the bubble diameter around 3.5mm.
However, as for the other conditions and measurements such as turbulence, it still needs
further investigation
(iii) Duct flow
According to the different duct geometries, several experimental databases of
the upward turbulent bubbly flows have been established in the small circular and large
circular pipes, and small non-circular and large noncircular ducts. The small and large
ducts were distinguished by the relation between the duct size DH and maximum bubble
size DMax ( gi
g 40 ) under the tested conditions, i.e., the small duct if DH <
DMax and otherwise large ducts [1.7]. The detailed literature survey of this flow type
could be found in [1.8]. In this section, the representative experimental studies and the
observations are briefly summarized.
In the small circular pipes with pipe diameter ranging from 15mm and 60mm,
the representative experimental studies were carried out by Serizawa [1.9-1.10],
Theofanous and Sullivan [1.11], Wang et al. [1.12], Liu [1.13-1.15], and Hibiki et al.
8
[1.16-1.17] by measuring the flow patterns and local flow distributions including
averaged liquid and gas velocities, void fraction, turbulence, bubble size, bubble
interfacial area concentration under several flow conditions. Therein, it was observed
that: for the upward turbulent bubbly flow in the small circular pipes, the pronounced
wall peak of the void fraction occurs for low area-averaged void fraction flow
conditions but core-peak occurs for the higher void fraction. The interesting
phenomenon was attributed to the lift force exerted on the bubbles which were
dependent on the bubble size [1.13]. With the wall peak of void fraction, the average
axial liquid velocity (Wl) profiles were observed to be more uniform in the core region
[1.9-1.17]. For higher void fraction cases, Theofanous and Sullivan [1.11] and Wang et
al. [1.12] observed slightly shift of the location of the maximum liquid velocity profile.
As for the turbulent structures, the radial distribution of the turbulence was observed to
be more uniform in the core region comparing to that of the single-phase flow.
Moreover, similar to the mirco-bubble case the turbulent drag reduction near the wall
was also observed under the low void fraction [1.18-1.19].
For the upward turbulent bubbly flow in the large diameter circular (pipe
diameter larger than 100 mm), Ohnuki and Akimoto [1.20] and Shawkat et al.
[1.21-1.22] established the comprehensive databases including the velocity field,
turbulent kinetic energy, void fraction, bubble size, bubble velocities and so on. The
experiments carried out [1.23-1.29] mainly measured the bubbles information including
void fraction, bubble size, bubble velocities and the interfacial area concentration
without the liquid phase flow field. According to the above databases, it was obtained
that the bubbles tend to migrate toward the pipe center earlier forming a core-peak of
the void fraction profile rather the wall-peak void distribution under the same flow
conditions even for the smaller bubbles with the bubble diameter 3-5mm [1.22], which
is different from those in small circular pipes. The wall-peak of the void fraction was
only observed for the cases with very low averaged void fraction less than 4% in the
experiments carried out by [1.20, 1.22, 1.24] in a 200mm pipe when the bubble diameter
was relatively small. No significant difference on the magnitude and shape for the
9
average liquid velocity was observed except the lower value in the near wall region
compared to that in the small circular pipe. Besides, as compared with that in the
smaller pipe, Prasser et al. [1.30] observed that under the same superficial velocities the
bubbles with larger diameter could move more freely and become more deformed in the
large pipes. The modifications of the distribution of the void fraction were attributed to
the deformation of the bubbles [1.31-1.33]. As for the liquid turbulence structure was
also considered to be one of the main mechanisms for the formation of the core-peak of
the void fractions in the large pipes. Lahey et al. [1.34] and Ohnuki and Akimoto [1.35]
suggested the strong turbulence diffusion effect on the bubbles is regarded as one reason
for the above phenomenon. However, more experimental databases of the liquid
turbulence are still necessary to validate these points.
For the upward turbulent bubbly flow in the small noncircular ducts, Lopez de
Bertodano et al. [1.36] established the comprehensive databases of the velocity fields,
Reynolds stress tensor and void fraction in the triangle duct. The other experiments in
the square and rectangular ducts by [1.37-1.41], mainly measured the void fraction.
According to the above databases, for noncircular ducts cases, it was observed that due
to the non-homogeneous geometrical boundary bubbles tend to be accumulated more in
the corner region, thus leading to the more significant corner peak of the void fraction
compared to the wall peak. The average liquid velocity was also found to be more
uniform comparing to the single-phase flow.
For the upward turbulent bubbly flow in the large noncircular ducts, recent
experiments done by Sun et al. [1.42] established detailed database for large vertical
square duct. Similar to the bubbly flow in other noncircular pipes [1.36-1.41], the
pronounced corner and wall peaks of void fraction were observed under the low
averaged void fraction cases. Due to the high local void fraction near the wall and the
corner the apparent peak of average liquid velocity was also found near corner and wall.
However, among the existing experiments in the circular small and large pipe and small
noncircular pipe no such strong velocity peak has been observed near the wall except
slightly peak near the wall [1.11-1.12]. In addition, the secondary flow in the cross
10
section observed in the single-phase turbulent flow in the noncircular ducts [1.43] may
have an effect on the flow characteristics. However, so far there are few experimental
measurements to show the existence and effect of the secondary flow.
1.2.2.1.2 Database of bubble behaviors
As mentioned above, the bubbles behavior is another key point to understand the
flow characteristics, the physical process and modeling of the turbulent bubbly flow in
the above flow types. So far, the basic databases of the bubble behaviors have been
established in the following flow types:
(i) Stagnant system
Therein, the important parameters including the terminal velocities VT, bubble
deformation E, and the drag coefficient CD were investigated based on the numerous
experiments of the single bubble [1.44]. The correlations of the above parameters were
proposed based on the following dimensionless numbers when it is dominated by
different effects such as the surface tension, inertial or viscosity effects.
2
Eol g BgD
,
4
2 3M
l l g
l
g
, Re l B R
B
l
D V
and
2
We l B RD V
. (1.1)
where Eötvös number Eo characterizes the relative importance of gravity to surface
tension; Morton number M the physical properties of systems; Reynolds number Re the
relative importance of inertial force to viscous force; and Weber number We the relative
importance of inertial force to surface tension. For the constant the fluid properties (the
constant M), the above dimensionless numbers solely depend on the bubble size DB,
since the relative velocity VR is also determined by the bubble size and the fluid
properties.
(ii) Uniform shear flow
Single-bubble experiments in the uniform shear flow were carried out by
Tomiyama et al.[1.33], where the bubble’s lateral migration was found to depend on the
11
bubble size DB or Eo. For example, the migration direction for the small bubble or Eo
was opposite to that for the large bubble or Eo. Besides, the well-known lift force
coefficient correlation was proposed, which will be given in the later section.
(iii) Upward turbulent bubbly flow
For the upward turbulent bubbly flow, Liu [1.13] studied the bubble size effect
on the flow characteristics in the small circular pipe by carefully controlling the inlet
bubble size DB. The lateral migration of bubbles and the flow regime transition were
found to be very sensitive to the bubble size. Under the low void fraction for the cases
of the bubble size DB < 5~6mm, the wall peak of the void fraction could be obtained,
while the core peak of the void fraction was observed for the large bubble size DB
>5~6mm. This was in consistent with the observation of Tomiyama’s single bubble
experiments [1.33]. In addition, the entrance length effect L/DH was also confirmed to
be an important parameter to determine the lateral void distribution due to the pressure
difference which could induce the bubble expansion. Along the axial direction, the
different flow regimes from the bubbly flow to slug flow are possible to be observed
due to the developing of the bubble coalescence and the bubble break-up.
1.2.2.1.3 Summary on experimental measurements
In summary, based on the above experimental databases, the following main
achievements were obtained for the turbulent bubbly flow:
(i) Bubble behaviors
For the upward turbulent bubbly flow in the ducts, there may exist the bubbles
lateral migration due to the additional force exerted on the bubbles from the liquid phase,
which could cause the phenomenon of the wall-peak, double peak or the core peak of
the void fraction. The lateral migration of the bubbles depends on the bubble size, the
bubble deformation, the duct size, and also the duct geometry. For the small bubbles,
under the low averaged void fraction the wall peak of the void fraction distribution
tends to be formed, while the core peak of the void fraction distribution tends to be
12
formed for the large bubbles. The duct size and geometry affects the void fraction under
which the transition from the wall peak to the core peak occurs earlier. Moreover,
increasing the averaged void fraction, the flow regime transition occurs which might
change from the dispersed bubbly flow to the annular flows, for which several flow
regime maps have been proposed [1.2].
(ii) Velocity fields in the liquid phase
The averaged liquid phase velocity and the turbulence fields were mainly
focused on. For the upward turbulent bubbly flow with the wall peak of the void
fraction, the liquid phase velocities were found to be more uniform in the core region.
For the turbulent bubbly flow with the core peak of the void fraction, the liquid phase
velocities were found to be steeper comparing to that of the single phase flow.
(iii) Turbulence modulation in the liquid phase
With the presence of the bubbles, the turbulence modulation occurs and
according to the above experiments, the observation could be summarized as follows:
(a) How the turbulence of the liquid phase was modulated strongly depends on
the bubble size and void fraction. Generally, with smaller particle or bubble
size smaller than the eddy size, the turbulence reduction occurs, while for
the large bubble size, the turbulence attenuation occurs [1.45]. However, for
the turbulent bubbly flow with relative large bubble size, under very low
void fraction cases such as less that 5%, the turbulence suppression could
also occur [1.19, 1.21]. For higher void fraction cases, the turbulence
enhancement occurs.
(b) The turbulent energy spectra in the turbulent bubbly flow were only
considered under very low void fraction. Therein, it was found that the
classical Kolmogorov power law (-5/3) was replaced by (-8/3) [1. 4, 1.21].
(c) Few detailed discussions of the turbulence anisotropy were carried out for
the upward turbulent bubbly flow, among which it was found to be
controversial. Theofanous and Sullivan showed the presence of the bubbles
13
didn’t alter the turbulent isotropy level [1.11]. Similar results were also
reported by Lance et al. [1.4, 1.5]. However, Wang et al. [1.12] showed the
bubbles modification of the local turbulence isotropy occurs which were
more anisotropic turbulence in the core region but more isotropic turbulence
near the wall [1.12].
1.2.2.2 Numerical simulations
Except experimental measurements, DNS has been proved to be a powerful tool
and played very important role in developing the turbulent models for the single-phase
turbulent flow. By directly solving the governing equations of the flow, DNS could
provide us all the scales of the turbulence without resorting any empirical closure
assumptions. Based on the VOF methods which are used to reconstruct the interface
structures, DNSs have been carried out for the turbulent flow with single bubble or
several bubble cases. However, for the cases with a large number of the bubbles, DNSs
started very late and are still very limited.
For the turbulent bubbly flow, currently Tryggvason group [1.46-1.56] has
constructed the numerical algorithms based on the finite difference method and the front
tracking of the bubble interface. With this numerical algorithm, a series of DNS studies
were performed for the homogeneous turbulent bubbly flow, vertical channel turbulent
bubbly flow and so on [1.46-1.56]. Therein, the detailed flow characteristics including
the flow fields, turbulence and bubbles behaviors were investigated.
For example, similar to the uniform turbulent bubbly flow carried out by Lance
and Bataille [1.4], Esmaeeli and Tryggvason [1.46-1.48] carried out DNSs of two- and
three- dimensional buoyant bubbles with regular array and freely array in the periodic
domains for the low and moderate Reynolds numbers. The turbulence generated therein
was compared the turbulence prediction based on the potential model [1.4]. The large
difference was found especially for the three-dimensional flow. In addition, the inverse
energy cascade of the turbulent structures was also noticed. Bunner and Tryggvason
[1.49-1.51] extended the previous DNS study [1.46-1.48] to more buoyant bubbles by
14
focusing on the bubbles behaviors including the rising velocity, the microstructures, the
dispersion and the fluctuation velocities of the bubbles and the flow fields such as the
bubble-induced pseudo-turbulence. Therein, the turbulent energy spectra showed a -3.6
power-law distribution at large wave numbers in consistent with Lance’s experimental
observations [1.4]. Moreover, the turbulence was found to be strongly anisotropic and
the anisotropy declined as the void fraction increased. The effects of the bubble
deformation on the flow properties were also investigated by Bunner and Tryggvason
[1.52], which showed a preference for pairs of the deformed bubbles to be vertically
aligned and for pairs of spherical bubbles to be aligned horizontal. Consequently, the
deformed bubbles tend to move toward the former bubbles’ wake, by which the stronger
velocity fluctuations could be induced by the deformed bubbles. Similar to the duct flow,
Lu et al. [1.53-1.55] carried out DNSs of the upward and downward turbulent bubbly
flows in a vertical channel with the low void fraction. Their numerical results [1.55]
showed that the bubble deformation rather than bubble size dominated the bubbles
lateral migration. In the upward flow, the deformed bubbles tend to migrate toward the
channel center, while the spherical bubbles tend to move the channel walls. Even with
relatively larger bubble size, the turbulent drag reduction was also obtained under low
void fraction when the bubble size was comparable to the buffer layer. Tryggvason
[1.56] addressed the multi-scale issues of the turbulent bubbly flow, and suggested the
multi-scale model for the turbulent bubbly flow based on the DNS databases.
Based on the DNS of the turbulent flow in a upward vertical channel, Bolotnov
[1.57] paid attention to the isotropy level of Reynolds stress distribution for the
single-phase and bubbly flow in order to address whether the isotropic turbulent
assumptions are suitable for the bubbly flows applications or not. His numerical results
showed that the presence of the bubbles improved the turbulence isotropy for low
Reynolds number flows, and didn’t alter for the moderate Reynolds number cases.
However, it is notable that the DNSs [1.57] only considered the extremely low void
fraction cases, say less than 3%.
15
1.2.2.3 Summary
So far, from the literatures, the numerical databases on the turbulent bubbly flow
have been established as summarized in Table 1.1, based on which our understanding on
the detailed fundamental processes and mechanisms of the turbulent bubbly flow were
improved. However, it is worth noting that in the above numerical simulations, only
tens at most hundreds of the bubbles were simulated due to the increasing of the
numerical difficulty especially the computer requirement, i.e., the numerical databases
were still limited in the low void fraction cases as compared to the real experiments.
There still exist some qualitatively controversial topics especially the bubble effects on
the turbulent anisotropy. In the simulations of the turbulent uniform bubbly flow by
Tryggvason groups [1.51-1.52], the turbulent behaves strongly anisotropic, while in
Bolotnov’s simulation in the vertical channel, the presence of the bubbles improve the
turbulence isotropy [1.57].
16
Table 1.1 Numerical databases based on DNSs of the turbulent bubbly flow.
Numerical Geometry Researchers Void fraction α
Bubble shape
S: Spherical;
D: Deformed
Note
2D Uniform flow Esmaeeli and Tryggvason
[1.46-1.48] 12.56% Eo = 1.0(S) -
3D Uniform flow
Esmaeeli and Tryggvason
[1.47-1.48]
Buner and Tryggvason
[1.49-1.52]
0%, 2%, 6%, 12%, 24%
Eo = 1.0(S)
Eo = 1.0(S), 5.0(D)
Regular bubble array
and Free bubbles
3D Vertical channel
Lu & Tryggvason
[1.53-1.55]
Bolotnov [1.57]
1.5%, 3.0%, 6.0%
0%, 1%
Eo = 0.45(S), 4.5(D)
Eo = 0.11(S)
-
-
17
1.2.3 Review of existing models for turbulent bubbly flow
Similar to the single-turbulent flow, though DNSs of the turbulent bubbly flow
could provide us the most detailed flow information, the requirements on the computer
memory and numerical time limited its application to the real industrial engineering
systems, especially for the large void fraction with a large number of the bubbles. The
models which could meet the requirements of the engineering application were more
efficient for the turbulent bubbly flows. This section will review the state-of-the-art of
the development and implementation of the models for the turbulent bubbly flows.
1.2.3.1 Model development
As mentioned before, in the turbulent bubbly flows the presence of the
multi-moving and deformable interfaces leads the flow to be significant discontinuous,
which brings the main numerical difficulties, i.e., resolving the fine interfacial structures.
During the model development of the turbulent bubbly flow the first important issue is
to eliminate the flow discontinuity and to obtain the macroscopic governing equations
which should satisfy the continuum fluid assumption based on the proper averaging
method.
Generally, according to the domain in the averaging method, the models for the
turbulent bubbly flow could be classified into two types, i.e., the one-dimensional
models and the multi-dimensional models. The one-dimensional models solve the
evolution of the cross-section averaged mean flow information such as the void fraction,
velocities, and pressure. The homogeneous equilibrium mixture model and the drift-flux
model by Zuber and Findlay [1.58] belong to this type. It requires less computer
memory and has been widely adopted in the past design calculations for the nuclear
reactor [1.59]. However, the models of this type require the prior knowledge of the local
distributions of phase, velocities and so on which are normally not known. As compared
to the one-dimensional model, the multi-dimensional models were derived based on the
time averaging, and can provide the local time-averaged flow information with the aid
of the powerful computer. Thus, to improve the models of the turbulent bubbly flow, the
18
numerical studies based on the multi-dimensional models are necessary.
According to the methods to consider the two phases during the averaging, the
models of the turbulent bubbly flow could be also classified into two types, i.e., the
mixture models and two-fluid models [1.2].
The mixture models consider the flow with both the continuous liquid phase and
the dispersed bubbles as a whole mixture, whose properties were determined by
coupling the properties of the two phases and each phase’s fraction. Consequently, the
conservation equations were established for the mass, momentum and energy based on
the whole mixture fluid. The homogeneous equilibrium mixture model is the simplest
example, which treats the fluid of the bubbly flow as a continuous mixture just like
single phase flow with modified the fluid properties. The relative motion between two
phases was neglected in the homogeneous equilibrium mixture model. Nowadays, the
drift-flux model proposed by Zuber and Findlay [1.58] is one widely used mixture
model, in which another equation for the relative motion between two phases was
included and the effects of two-phase interaction on the flow could be considered.
The two-fluid model proposed by Ishii [1.60] treats the two phases separately.
For each phase, the conservation equations including the mass, momentum and energy
were established based on the time averaging. Though the two-fluid model is more
complicated as compared with drift-flux model, however, in this way the more detailed
flow information such as the transfer processes of the mass, momentum and energy
between two phases could be predicted. Nowadays, the two-fluid model has been paid
lots of attentions, based on which the numerical predictions of the turbulent bubbly flow
were improved, and in the present dissertation this model will also be considered.
The basic governing equations of two-fluid model were shown in Eqs. (1.2) and
(1.3). Here, the flows were restricted to the adiabatic and incompressible two-phase
flows without the phase change.
0 kkkk
Dt
DV
(1.2)
19
kkkikkiwkik
kkkkkkkkkk
kkkk
pp
pDt
D
τMM
gvvVV
(1.3)
where the subscript-k refers to the liquid (l) and gas (g) flow; the subscript-i refers to the
effect of the interfaces; the overbar means the time-averaged quantities; the bold
variables are the vectors or tensors; k is void fraction of phase-k; Vk is the
corresponding phase velocity vector and could be separated by the mean value and the
fluctuations kkk vVV ; vk the fluctuating velocity vector; Dk /Dt indicates the
material derivative; ρk and μk are the density and viscosity of the corresponding phase-k;
pk is the static pressure of the corresponding phase-k; g is the gradational acceleration;
Mik is the interfacial force on phase-k; Mwk is the wall shear force on phase-k;
kkkkkk vvVτ ; kkk vv is Reynolds stress in the phase-k.
Comparing to the single-phase turbulent flow, the averaging process generates
the new interfacial transfer terms as shown by the last four terms in Eq. (1.3). The
interfacial term Mik related with the jump condition stratify Mil+ Mig = 0. In addition,
the turbulent Reynolds stresses also includes the phase interaction through the interfaces.
To improve the model of the turbulent bubbly flow the current work was mainly focused
on the establishment and validation of the constitutive relations for the interfacial
transfer terms and the turbulent Reynolds stresses based on the established experimental
databases. Some attentions were also paid to the numerical algorithms to solve the
above equations. In the following, the state-of-the-art of the constitutive relations for the
two-fluid model will be reviewed in brief.
1.2.3.1.1 Interfacial transfer terms
During the model development, the interfacial term Mik was decomposed into
the drag force d
ikM and the non-drag forcend
ikM ,
nd
ik
d
ikik MMM (1.4)
20
The drag force d
ikM is generally expressed by the following standard form in
terms of the drag coefficient CD, the relative velocity Vr between two-phases and the
interfacial area concentration ai of the bubbles [1.60]:
irrlD
d
ig aC VVM 21 . (1.5)
The attentions were paid to establish the correlations for CD and ai. Several
correlations for CD have been established based on the experiments of the single solid
particle, droplets, or bubbles in the infinite systems [1.44]. Ishii and Chawla [1.61]
developed the correlation of CD for the multi-particle systems considering the flow
regimes and the characteristics of the particles.
As for the interfacial area concentration ai, the simplest expression is given in
terms of the Sauter mean diameter DSM and the void fraction α assuming the bubbles to
be spherical shape: ai=6/DSM, where DSM is the volume equivalent bubble diameter of
spheres. Ishii and Mishima [1.62] developed the correlations of the interfacial area
concentration ai considering the flow regimes as shown in Eq. (1.6).
6
64.5 11 1
For bubbly flow
For slug or churn turbulent flow
SM
gs gs
H gs H gs
D
i
D D
a
. (1.6)
As for the non-drag forcend
ikM , the following forces were usually considered
according to the single-particle theory.
Lahey et al. [1.63] studied the constitutive equations of the virtual mass force for
the bubbly flow. They found that the introduction of the virtual mass force V
igM term
could improve the numerical stability and efficiency, and thus was important to the
two-fluid model. Ishii and Mishima [1.62] summarized and gave the correlations for
V
igM in term of the virtual mass coefficient CVM and the acceleration of the relative
velocities according to the flow regimes as
V rig VM l r r
DC
Dt
VM V V , (1.7)
21
and
12
1 2 For bubbly flow
1
1 /5 0.66 0.34 For slug flow
1 / 3
VM
b b
b b
CD L
D L
(1.8)
According to the analysis of the single-spherical particle moving in the viscous
shear flow [1.44], the particle could receive the lift force L
igM from the liquid phase
perpendicular to the flow direction. Similarly, for the turbulent bubbly flow with the
relative velocity Vr and the liquid phase velocity gradient lV , the lift force exerted on
the bubbles was also considered and expressed in term of the lift force coefficient CLM,
the relative velocity Vr, and the liquid phase velocity gradient lV as
L
ig L l r lC M V V . (1.9)
The lift force played a very important role to the prediction of the bubble lateral
distributions [1.10-1.17]. Attentions have been paid to determine the lift force
coefficient CLM based on the single bubble experiments and simulations [1.33, 1.34,
1.44], which was found to be related with the bubble size and deformation. According to
Tomiyama’s experiments [1.33], the lift force coefficient CLM was given as follows:
H
H
min 0.288tanh 0.121Re , Eo For Eo 4
Eo For 4 Eo 10
0.27 For 10 E
H
LM H
f
C f
Ho
(1.10)
where 3 2
H H H HEo 0.00105Eo -0.0159Eo -0.0204Eo 0.474f ; 2
HEoj g HBg D
;
HBD the maximum horizontal bubble dimension.
The constitutive relations of the turbulent dispersion force T
igM exerted on the
bubbles were derived by Lahey [1.34], Bertodano [1.65-1.66], and Burns [1.67]
respectively using different methods as follows:
22
13
4 0.9 1
From Lahey [1993]
From Bertodano [2006]
From Burns [2004]
t
b
t
T l
T
ig T l
T D r ld
C k
C
C C
M
V
(1.11)
where CT is the turbulent diffusion coefficient; k is the turbulent kinetic energy; υt is the
eddy viscosity.
Antal et al. [1.68] introduced and derived the wall-lubrication force w
igM
exerted on the bubble into the two-fluid model to prevent the bubble from touching the
wall based on the analysis of single spherical bubble moving in the laminar flow. The
lateral distribution near the wall could be better predicted with this force w
igM as
expressed in the following form:
22w
ig l W r wdCM V n (1.12)
where wn is the wall-normal unit vector; CW is the wall force coefficient which is
related with the distance y to wall and expressed as
3 3
2 2
2
4 4
2
0.104 0.06 0.147 From Antal [1991]
Eo From Tomiyama [1998]
0.005 0.05 From ANSYS [2010]
B
B B
H
B
D
r w y
D D
W y D y
D
y
C y F
V n
(1.13)
with H
H
exp -0.933Eo 0.179 For 1 Eo 5Eo
0.007Eo 0.04 For 5 Eo 33F
.
The last two terms in Eq. (1.3) indicate the interfacial pressure and shear stress
terms. Lahey [1.70] and Bertodano et al. [1.71] considered the same pressure in both
phases pk=p, τik=τk and introduced the following constitutive relations for interfacial
pressure and shear stress terms:
23
2P
ig ig g p l rp p C M V (1.14)
S
ig ig g M τ τ (1.15)
where Cp is the pressure coefficient and 0.25 for non-interacting spherical bubbles based
on the inviscid result. Bertodano et al. [1.66] discussed the non-spherical cases where
Cp > 0.25 due to the wake effect behind the bubbles.
Though several experimental and analytical efforts were carried out to improve
the above correlation coefficients [1.2], so far the physical mechanism of the above
forces are still poorly understood. In addition, it is also notable that the aforementioned
correlations were mainly proposed based on the single particle case, for the bubbly flow
with many bubbles and bubble-bubble interaction further efforts are still necessary.
1.2.3.1.2 Turbulent Reynolds stresses
Turbulence term T
k k k kτ v v in the above momentum equations is one of the
most important key terms for the turbulent bubbly flow, which determines the prediction
of the distributions of the velocities, void fraction, and interfacial term. Turbulent
models for the single-phase turbulence have been widely studied and could be found
everywhere in the textbooks about the turbulence [1.1]. However, for turbulent bubbly
flow, the understanding and modeling of this term are still in an initial stage due to the
complicated physical process. Therein, the following mechanisms for the turbulence
generation coexist such as the general shear-induced turbulence, the interaction between
the shear-induced turbulent structures and the bubbles, the bubbles-induced turbulence
and its interaction between the different turbulent structures [1.5, 1.9]. So far, the
development of the turbulent models for the turbulent bubbly flow was mainly carried
out based on the analogy of the single-phase turbulent models.
Based on the phenomenological modeling where the eddy viscous assumption
for the turbulent shear stress / 2T T
S k l t l l τ V V was also adopted, Sato
and Sadatomi [1.72], Drew and Lahey [1.73] and Kataoka and Serizawa [1.74] adopted
24
the zero or one equation model to the turbulent bubbly flow to determine the turbulent
shear stress by adding the bubble-induced turbulence. Therein, the attentions were also
paid to the determination of the eddy viscosity t and the mixing length lm.
2
From Sato, 1981
From Drew and Lahey, 1981
2 From Kataoka and Serizawa, 1995
3
l s b
T
S l m
l s B
dV
dr
dV dVl
dr dr
k dVl l
dr
(1.16)
where s and b are the eddy viscosity induced by shear and bubbles, in Sato and
Sadatomi [1.72] b b B rC D V ; sl and bl are the mixing lengths of the shear
induced turbulence and bubble induced turbulence respectively; s and sl were
decided based on the Prandtl mixing theory. However, the determinations of the eddy
viscosity and mixing length are strongly dependent on their experimental data, and so
far the applications of the above models were still limited and need further validation.
Theofanous and Sullivan [1.11] and Wijngaarden [1.75] theoretically and
statistically discussed the bubble-induced turbulence or the pseudo-turbulence by
comparing the turbulent intensities before and after the bubbles injection. Based on the
potential theory Wijngaarden [1.75] derived the pseudo turbulent Reynolds stress for the
rising bubbles in the laminar flow as
23 1
20 20B r r rI
τ V V V , (1.17)
and for the uniform turbulent flow,
2
22 0
26 0rE
r
uk M f f a
VV
(1.18)
For the bubbly flow in the pipe, Theofanous and Sullivan [1.11] estimated the
total turbulence based on the mean flow velocity and the wall shear stress as
25
2
11 1 1 1 1 1.5
2 4
g H Hf
c l c B
vv D g Dc
V V D
. (1.19)
Several more general and accurate turbulent models for turbulent bubbly flow
have also been developed on basis of the two-equation model or Reynolds stress model
of the single phase turbulent flow [1.1]. Kataoka et al. [1.76] derived the basic
equations for the turbulent Reynolds stress and dissipation for the two-phase flow,
where the expressions of the interfacial terms were given and discussed. So far,
according to the author’s knowledge, the following three different methods were used to
consider the turbulence.
Lahey [1.70] developed η-ε model for the turbulent stresses in the liquid phase of
the turbulent bubbly flow by mimicking the single-phase η-ε model as follows:
2l l l S l i
D kC
Dt
vv vv vv P Φ I S (1.20)
1 2i
l l l l B
e
D SC C P C C
Dt k k k t
vv (1.21)
P, Φ, ε and I are the turbulence production, redistribution, dissipation and unit
tensor which are the same as the single-phase turbulent flow. Cε, Cε1, Cε2 and CεB are the
model constants. For the turbulent bubbly flow, the additional interfacial tensor Si and
/B i eC S t are the bubble induced turbulence generation and dissipation;
tracei iS S ; te is the characteristic time of the eddies, and the choice of te was
discussed by Rzehak and Krepper [1.77].
By 12
tracek vv , the above Reynolds stress model become the k-ε model for
the turbulent bubbly flow in the following form:
l l l t l i
Dk k P S
Dt . (1.22)
Several studies were paid to the constitutive relations of the additional interfacial
26
tensor Si and the additional interfacial dissipation term /B i eC S t [1.77].
Bertodano et al. [1.71] proposed another type of η-ε model considering the
transport equations for both the shear-induced turbulent and the bubble-induced
turbulence:
l S l l t S l S S
Dk k P
Dt . (1.23)
BBam
b
lBtllBl kk
tkk
Dt
D
. (1.24)
1 2
l t S Sl S l S l S
t
DC P C
Dt k k
. (1.25)
In addition, the turbulent shear stress was expressed as
2
/ 23
T
l t S l S B Bk k τ V V A A (1.26)
where the eddy viscosity
2
t b B r
kC C D
V , and AS and AB are the tensors to
show the turbulence anisotropy for the sear-induced turbulence and bubble-induced
turbulence respectively.
Similar to Lahey [1.70] and Bertodano et al. [1.71], Chahed et al. [1.78-1.79]
proposed another turbulent model by decomposing the turbulent Reynolds stress into
turbulent dissipative part due to the velocity gradient and the pseudo-turbulent
non-dissipative part due to the interfacial drag. Therein, the turbulent anisotropy was
considered by introducing the bubble characteristics time following the experimental
results and initial model of Lance [1.4-1.6].
On the other hand, as observed in the experimental and numerical databases, the
bubbles size and deformation could strongly affect the flow characteristics. The
aforementioned models were developed based on a prior knowledge of the bubbles size
and deformation such as the averaged bubble Sauter mean diameter DSM. However, in
the real systems, there exist a broad range of the bubble sizes, and the bubbles could
27
experience the coalescence with other bubbles and also breakup, which were not shown
in the above models at all. Recently, to improve the models for the turbulent bubbly
flow, the attentions were also paid to the bubble behaviors such as distributions of the
bubble size or interfacial area concentration (IAC) [1.2]. Generally, the models to
predict the bubble behaviors are based on Boltzman statistical averaging method and
Boltzman equation for the dispersed bubbles. Correspondingly, the dispersed gas phase
was separated into several groups according to the interested properties of the dispersed
bubbles such as bubble size and IAC. For each group of the dispersed phase, the
conservations equations of the mass, momentum, and energy were established
considering the birth and death of the each group. For the continuous phase, the
interfacial effects from the dispersed phase are the sum of all groups. With this method,
it aimed at predicting the behaviors of the dispersed gas phase such as the distributions
of the bubble size or interfacial area concentration which are crucial to the flow
characteristics. Currently, the population balance model [1.80], the one-group and
two-group interfacial transport model [1.2] and multi-fluid model [1.81] belong to the
model of this type. Therein, the attentions were mainly paid to model the birth and death
of the each group which depend on the bubble coalescence and breakup. The
corresponding detailed reviews could be found in [1.82-1.83]. Due to the complicated
nature of the bubble coalescence and breakup, these models are still in an initial stage
and the more efforts are required in the future work.
1.2.3.2 Model implementation and validation
As mentioned in Section 1.2.2, the current experimental databases of the
turbulent bubbly flow were mainly established in the following flow types: uniform
bubbly flow, bubbly boundary layer flow, the bubbly flow in the small and large circular
pipes and small and large noncircular ducts. With the aid of these databases, the
implementations and validations of the two-fluid model were carried out such as in
[1.70-1.85]. At this stage, most of the attentions were focused on the prediction of the
phase distributions by considering the interfacial terms in the momentum transport
28
equations [1.70-1.74]. However, for the prediction of the characteristics of the turbulent
structures such as intensities and anisotropic properties, even several models have been
proposed [1.75, 1.78-1.79], they were limited for very low void fraction or not
satisfactory due to our limited knowledge on the bubble-induced turbulence. Moreover,
the previous implementation and validation of models were mainly carried out in the
small circular pipes [1.70-1.71, 1.84]. For the bubbly flow in the large circular and
noncircular ducts where the flow were strongly affected by the duct geometry and
showed different characteristics, there were still very few detailed validation work
[1.34-1.35, 1.85]. In addition, with the above models only the drag enhancement could
be predicted due to the superposition of the shear-induced turbulence and
bubble-induced turbulence, whereas the turbulent drag reduction were also observed in
the experiments [1.19]. Therefore, whether the models proposed and validate before are
suitable for the different conditions still needs further studies based on the new
experimental databases. And the new models for other flow properties such as
turbulence are also required.
1.2.4 Summary on review of databases and models
In section 1.2.2-1.2.3, the state-of-the-art of the turbulent bubbly flow was
briefly reviewed including the existing databases and models. So far, the experimental
databases of the turbulent bubbly flow have been developed for several flow types
including uniform bubbly flow, circular pipe flow, and noncircular duct flow. The flow
characteristics such as phase distribution and turbulence were mainly analyzed and
modeled for the turbulent bubbly flow in the small circular pipes and noncircular ducts.
However, for the turbulent bubbly flow in the large and non-circular ducts which could
be often met in the engineering applications, it is still lack of the experimental database
and the model evaluations. The previous developed models were focused on the
prediction of the phase distributions, but about the turbulent characteristics such as
intensities, turbulence generation process and anisotropic properties the current
understandings are still far from the satisfactory and limited for very low void fraction.
29
1.3 Scope and layout of the dissertation
On basis of the background of the turbulent bubbly flow mentioned above, the
present dissertation is aimed at the following aspects to improve the understanding and
modeling of the turbulent bubbly flow especially in the large noncircular duct:
(1) Evaluating the models proposed for the turbulent bubbly flow in the small
circular and noncircular ducts to that in the large noncircular duct;
(2) Further analyzing the flow characteristics of the turbulent bubbly flow in
the large noncircular ducts especially the liquid phase velocities and
turbulence;
(3) Improving the current models for the prediction of the turbulent bubbly
flow in the large noncircular ducts especially for the turbulence structures
based on the proposed analysis above.
The dissertation is composed by 8 chapters and arranged as follows: Chapter 1
mainly introduces the motivation, the current research status and the purpose of this
study. Chapter 2 describes the experimental database including the experimental
apparatus, flow conditions and some experimental results. Chapter 3 describes the
numerical methodologies based on the two-fluid model for the turbulent bubbly flow in
the large square duct, numerical results and discussions. To improve the modeling of the
turbulent bubbly flow in the large noncircular ducts, the further analysis and modeling
improvement are carried out for flow characteristics in the later four chapters. Chapter 4
discusses the developing process in the turbulent bubbly flow. Chapter 5 is focused on
the characteristics and modeling of the liquid phase velocities. Chapter 6 proposes a
new correlation for the bubble deformation and describes the numerical validation of the
void fraction. Chapter 7 describes the turbulence modulation due to the presence of the
bubbles. Finally, the summary of this dissertation and the recommendation of the future
work will be given in Chapter 8.
30
Reference
[1.1] Pope, S. B., 2000, Turbulent flows, Cambridge University Press.
[1.2] Ishii, M., Hibiki, T., 2011, Thermo-fluid dynamics of two-phase flow. 2nd ed.
New York: Springer.
[1.3] Marié, J. L., 1983, Investigation of two-phase bubbly flows using Laser Doppler
Anemometry, PCH Physico Chemical Hydrodynamics 4, 103-118.
[1.4] Lance, M. , Bataille, J., 1991, Turbulence in the liquid phase of a uniform
bubbly air-water flow, J. Fluid Mechanics 222, 95-118.
[1.5] Lance, M. , Bataille, J., 1991, Homogeneous turbulence in bubbly flows, J.
Fluids, Engineerig 113, 295-300.
[1.6] Moursali, E., Bataille, J., Marié, J. L., 1995, An upward turbulent bubbly
boundary layer along a vertical flat plate, Int. J. Multiphase Flow 21, 107-117.
[1.7] Isao, K., Mamoru, I., 1987, Drift flux model for large diameter pipe and new
correlation for pool void fraction, Int. J. Heat Mass Transfer 30, 1927–1939.
[1.8] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
[1.9] Serizawa, A., 1974, Fluid dynamics characteristics of two-phase flow, Doctor
Thesis, Kyoto University.
[1.10] Serizawa, A., Kataoka, I., Michiyoshi, I., 1975, Turbulence structure of air -
water bubbly flow—II. local properties, Int. J. Multiphase Flow 2, 235–246.
[1.11] Theofanous, T. G., Sullivan, J., 1982, Turbulence in two-phase dispersed flows,
J. Fluid Mechanics 116, 343-362.
[1.12] Wang, S., Lee, S., Jones, O., Lahey, R., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
[1.13] Liu, T. J., 1993, Bubble size and entrance length effects on void development in
a vertical channel, Int. J. Multiphase Flow 19, 99–113.
[1.14] Liu, T. J., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
31
pipe—I. liquid mean velocity and turbulence measurements, Int. J. Heat Mass
Transfer 36, 1049–1060.
[1.15] Liu, T. J., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
pipe—I. bubble void fraction, bubble velocity and bubble size distribution, Int. J.
Heat Mass Transfer 36, 1061–1072.
[1.16] Hibiki, T., Ishii, M., 1999, Experimental study on interfacial area transport in
bubbly two-phase flows, In. J. Heat Mass Transfer 42, 3019–3035.
[1.17] Hibiki, T., Ishii, M., Xiao Z., 2001, Axial interfacial area transport of vertical
bubbly flows, In. J. Heat Mass Transfer 44, 1869–1888.
[1.18] Michiyoshi, I., Serizawa, A., 1986, Turbulence in two-phase bubble flow, Nucl.
Eng. Des. 95, 253-267
[1.19] Serizawa, A., Kataoka, I., 1990, Turbulence suppression in bubbly two-phase
flow, Nucl. Eng. Des. 122, 1–16.
[1.20] Ohnuki, A., Akimoto, H., 2000, Experimental study on transition of flow pattern
and phase distribution in upward air–water two-phase flow along a large vertical
pipe, Int. J. Multiphase Flow. 26, 367–386.
[1.21] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2007, On the liquid turbulence
energy spectra in two-phase bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 33, 300-316.
[1.22] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2008, Bubble and liquid turbulence
characteristics of bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 34, 767-785.
[1.23] Sun, X., Smith, T. R., Kim, S., Ishii, M., Uhle, J., 2003, Interfacial structure of
air-water flow in a relatively large pipe. Exp. Fluids 34, 206-219.
[1.24] Shoukri, M., Hassan, I., Gerges, I., 2003, Two-phase bubbly flow structure in
large-diameter vertical pipes, The Canadian J. Chem. Eng. 1 205-211.
[1.25] Shen, X., Mishima, K., Nakamura, H., 2005, Two-phase phase distribution in a
vertical large diameter pipe, Int. J. Heat Mass Transfer 48, 211–225.
32
[1.26] Schlegel, J.P., Sawant, P., Paranjape, S., Ozar, B., Hibiki, T., Ishii, M., 2009,
Void fraction and flow regime in adiabatic upward two-phase flow in large
diameter vertical pipes, Nucl. Eng. Des. 239, 2864–2874.
[1.27] Smith, T.R., Schlegel, J.P., Hibiki, T., Ishii, M., 2012, Two-phase flow structure
in large diameter pipes, Int. J. Heat Fluid Flow 33, 156–167.
[1.28] Shen, X., Hibiki, T., Nakamura, H., Developing structure of two-phase flow in a
large diameter pipe at low liquid flow rate, Int. J. Heat Fluid Flow. 34 (2012)
70–84.
[1.29] Schlegel, J.P., Miwa, S., Chen, S., Hibiki, T., Ishii, M., 2012, Experimental
study of two-phase flow structure in large diameter pipes, Exp. Thermal Fluid
Sci. 4, 12–22.
[1.30] Prasser, H. M., Beyer, M., Boettger, A., Carl, H., Lucas, D., Schaffrath, A.,
Schuetz, P., Weiss, F. P., Zschau, J., 2005, Influence of the Pipe Diameter on the
Structure of the Gas-Liquid Interface in a Vertical Two-Phase Pipe Flow, Nucl.
Tech. 152, 3-22
[1.31] Esmaeeli, E. A., Tryggvason, G., 1994, Numerical simulations of rising bubbles
in bubbles dynamics and interface phenomena, Kluwer Academic, Dordrecht.
[1.32] Ervin, E. A., Tryggvason, G., 1997, The rise of bubbles in a vertical shear flow.
J. Fluids Eng. 119, 443-449.
[1.33] Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002, Transverse migration of
single bubbles in simple shear flows, Chem. Eng. Sci. 57, 1849-1858.
[1.34] Lahey, Jr R. T., Bertodano, de M. L., Jones, Jr O. C., 1993, Phase distribution in
complex geometry conduits, Nucl. Eng, Des. 141, 177-201.
[1.35] Ohnuki, A., Akimoto, H., 2001, Model development for bubble turbulent
diffusion and bubble diameter in large vertical pipes. J. Nucl. Sci. Technol. 38,
1074-1080.
[1.36] Bertodano, de M. L., 1992, Turbulent bubbly two-phase flow in a triangular duct.
Doctoral Thesis, Resselaer Polytechnic Institute.
33
[1.37] Launder, B.E., Ying, W.M., 2006, Secondary flows in ducts of square
cross-section, J. Fluid Mech. 54, 289.
[1.38] Ohba, K., Yuhara, T., 1982, Study on vertical bubbly flows using laser doppler
measurements: 2nd report, turbulence structure in square duct flow. Trans. Japan
Society of Mech. Eng. 48, 78-87
[1.39] Ohba, K., Yuhara, T., 1985, Matsuyama H. Study on vertical bubbly flows using
laser doppler measurements: 3rd report, simultaneous measurement of bubble
and liquid. Trans. Japan Society Mech. Eng. 51, 4194-4200.
[1.40] Zun, I., 1990, The mechanism of bubble non-homogeneous distribution in
two-phase shear flow. Nucl. Eng. Des. 118, 155-162.
[1.41] Sadatomi, M., Sato, Y., Saruwatari, S., 1982, Two-phase flow in vertical
noncircular channels. Int. J. Multiphase Flow 8, 641–655.
[1.42] Sun, H. M., Kunugi, T., Shen, X. Z., Wu, D. Z. and Nakamura, H., Upward
air-water bubbly flow characteristics in a vertical square duct, J. Nucl. Sci. Tech.,
Vol. 51, 267–281(2014).
[1.43] Hoagland, L. L., 1960, Fully Developed Turbulent Flow in Straight Rectangular
Ducts - Secondary Flow, Its Cause and Effect on the Primary Flow, Ph.D. thesis,
Department of Mechanical Engineering, MIT.
[1.44] Clift, R., Grace, J.R., Weber M.E., 1972, Bubbles, drops, and particles, Science.
[1.45] Gore, R. A., Crowe, C. T., 1989, Effect of particle size on modulating turbulent
intensity, Int. J. Multiphase Flow 15, 279-285.
[1.46] Esmaeeli, A., Tryggvason, G., 1996, An inverse energy cascade in
two-dimensional low Reynolds number bubbly flows, J. Fluid Mech. 314,
315-330
[1.47] Esmaeeli, A., Tryggvason, G., 1998, Direct numerical simulations of bubbly
flows Part 1. Low Reynolds number arrays, J. Fluid Mech. 377, 313-345.
[1.48] Esmaeeli, A., Tryggvason, G., 1999, Direct numerical simulations of bubbly
flows Part 1. Moderate Reynolds number arrays, J. Fluid Mech. 385, 325-358.
34
[1.49] Bunner, B., Tryggvason, G., 1999, Direct numerical simulations of
three-dimensional bubbly flows, Phys. Fluids 11, 1967-1969.
[1.50] Bunner, B., Tryggvason, G., 2002, Dynamics of homogeneous bubbly flows
Part 2: Rise velocity and microstructure of the bubbles, J. Fluid Mech. 466,
17-52.
[1.51] Bunner, B., Tryggvason, G., 2002, Dynamics of homogeneous bubbly flows
Part 2: velocity fluctuations, J. Fluid Mech. 466, 53-84.
[1.52] Bunner, B., Tryggvason, G., 2003, Effect of bubble deformation on the
properties of bubbly flows, J. Fluid Mech. 495, 77-118.
[1.53] Lu, J., Fernández, A., Tryggvason, G., 2005, The effects of bubbles on the wall
drag in a turbulent channel flow, Phys. Fluids 17, 095102.
[1.54] Lu, J., Tryggvason, G., 2006, Numerical study of turbulent bubbly downflows in
a vertical channel, Phys., Fluids 18, 103302.
[1.55] Lu, J., Tryggvason, G., 2008, Effect of bubble deformability in turbulent bubbly
upflow in a vertical channels, Phys. Fluids 20, 040701.
[1.56] Tryggvason, G., 2013, Multiscale issues in DNS of multiphase flows, 8th
International Conference on Multiphase Flow, 2013, Jeju, Korea.
[1.57] Bolotnov, Igor A., 2013, Influence of bubbles on the turbulence anisotropy, J.
Fluids Eng. 135, 051301.
[1.58] Zuber, N., Findlay, J. A., 1965, Average volumetric concentration in two-phase
flow systems, J. Heat Transfer, 453-468.
[1.59] Wheeler, C. L., 1986, COBRA-NC: a thermal-hydraulic code of transient
analysis of nuclear reactor components: equations and constitutive models.
NUREG/CCR-3262.
[1.60] Ishii, M., 1975, Thermo-fluid Dynamic Theory of Two-phase Flow, Eyrolles.
[1.61] Ishii, M., Chawla, T. C., Local drag laws in dispersed two-phase flows, Argonne
Nation Lab Report, ANL-79-105.
[1.62] Ishii, M., Mishima, K., 1984, Two-Fluid model and hydrodynamics constitutive
relations, Nucl. Eng. Des. 82, 107-126.
35
[1.63] Lahey, Jr. R.T., Cheng, L. Y., Drew, D. A.,, Flaherty, J. E., 1980, The effects of
virtual mass on the numerical stability of accelerating two-phase flows, Int. J.
Multiphase flow 6, 281-294.
[1.64] Tomiyama, A., Sou, A., Zun, I., Kanami, N., Sakaguchi, T., 1995, Effects of
Eötvös number and dimensionless liquid volumetric flux on local motion of a
bubble in a laminar duct flow, Advance In Multiphase Flow, Elsevier, 3-15.
[1.65] Bertodano, Lopez M., 1998, Two fluid model for two-phase turbulent jets, Nucl.
Eng. Des. 179, 65-74.
[1.66] Bertodano, Lopez M., Sun, X., Ishii, M., Ulke, A., 2006, Phase distribution in
the cap bubble regime in a duct, J. Fluids Eng. 977-1001.
[1.67] Burns, D., Frank, T., Hamill, I., Shi, J. M., 2004, The Favre averaged drag
model for turbulent dispersion in Eulerian multi-phase flow. In: 5th International
Conference on Multiphase Flow, Yokohama, Japan, 392.
[1.68] Antal, S.P., Lahey, Jr R. T., Flaherty, J. E., 1991, Analysis of phase distribution
in fully developed laminar bubbly two-phase flow, Int. Multiphase Flow 13,
309-326.
[1.69] Tomiyama, A., 1998, Struggle with computational bubble dynamics. In: 3rd
International Conference on Multiphase Flow, ICMF’98, Lyon, France.
[1.70] Lahey Jr. R.T., 1990, The analysis of phase separation and phase distribution
phenomena using two-fluid models, Nucl. Eng. Des. 122, 17-40.
[1.71] Bertodano, Lopez M., Lahey, Jr. R.T., Jones, O. C., 1994, Phase distribution in
bubbly two-phase flow in vertical ducts, Int. J. Multiphase Flow 20, 805-818.
[1.72] Sato, Y., Sadatomi, M., 1981, Momentum and heat transfer in two-phase bubble
flow—I. Theory, Int. J. Multiphase Flow 7, 167-177.
[1.73] Drew, D. A., Lahey, Jr. R.T., 1982, Phase distribution mechanisms in turbulent
low-quality two-phase flow in a circular pipe, J. Fluid Mech. 117, 91-106.
[1.74] Kataoka, I., Serizawa, A., 1995, Modeling and prediction of turbulence in
bubbly two-phase flow, Proc. 2nd
Int. Conf. Multiphase Flow, Kyoto,
MO2:11-16
36
[1.75] Wijngaarden, van L., 1998, On pseudo turbulence, Theoret. Comput. Fluid
Dynamics 10, 449-458.
[1.76] Kataoka, I., Serizawa, A., 1989, Basic equations of turbulence in gas-liquid
two-phase flow, Int. Multiphase Flow 5, 843-855.
[1.77] Rzehak, R., Krepper, E., 2013, Bubble-induced turbulence: Comparison of CFD
models, Nucl. Eng. Des. 258, 57-65.
[1.78] Chahed, J., Roig, V., Masbernat, L., 2003, Eulerian-Eulerian two-fluid model for
turbulent gas-liquid bubbly flows, Int. J. Multiphase Flow 29, 23-49.
[1.79] Bellakhel, G., Chahed, J., Masbernat, L., 2004, Analysis of the turbulence
statistics and anisotropy in homogeneous shear bubbly flow using a turbulent
viscosity model, Journal of Turbulence 5, 036
[1.80] Das, A. K., Das, P. K., 2010, Modeling bubbly flow and its transitions in vertical
annuli using population balance technique, Int. J. Heat Fluid Flow 31, 101-104.
[1.81] Politano, M. S., Carrica, P. M., Converti, J., 2003, A model for turbulent
polydisperse two-phase flow in vertical channels, Int. J. Multiphase Flow 19,
1153-1182.
[1.82] Liao, Y., Lucas, D., 2009, A literature review of theoretical models for drop and
bubble breakup in turbulent dispersions, Chem. Eng. Sci. 64, 3389-3406.
[1.83] Liao, Y., Lucas, D., 2010, A literature review on mechanisms and models for the
coalescence process of fluid particles, Chem. Eng. Sci. 65, 2851-2864.
[1.84] Bertodano, M. L., Lahey Jr, R. T. and Jones Jr, O. C., 1994, Development of a
k-ε model for bubbly two-phase flow. J. Fluids Engineering 116, 128-134.
[1.85] Lucas, D., Tomiyama, A., 2011, On the lateral lift force in poly-dispersed
bubbly flows, Int. J. Multiphase Flow 37, 1178-1190.
37
Chapter 2
Experimental databases of upward turbulent bubbly
flow
2.1 Introduction
Before evaluating the existed models of the turbulent bubbly flow in the large
noncircular ducts, a comprehensive database is necessary to be collected. Table 2.1
summarized the database information according to its duct geometry, which will be used
in the present dissertation. Sun’s experimental result in the large noncircular ducts
[2.10] will be chosen as the main database in the following chapters. In the following,
the experimental database will be reviewed and collected including the experimental
techniques and apparatus, the flow conditions and the experimental results.
2.2 Experimental techniques
To measure the velocity and turbulence fields, nowadays several experimental
techniques have been developed and could be classified into contact type and
contactless type according to whether inserting the probes into the flow or not. The
contact type techniques insert the probes into the flow and thus there may exist some
effects of the inserted probes on the flow. Therefore, nowadays for the single-phase
turbulent flow measurement, the contactless measurement techniques such as laser
Doppler velocimetry (LDV), particle tracking velocimetry (PTV) and the particle image
velocimetry (PIV) played more and more important role in the understanding and
modeling of the single-phase turbulence. Generally, the above contactless techniques are
based on the optical techniques and require the transparent fluids property. However,
unlike the single-phase turbulent flow, due to the low visibility the above optical
38
measurement techniques failed for the turbulent bubbly flow except for the cases with
the few bubbles and in the near-wall region. The velocities and turbulence in the
turbulent bubbly flow were mainly measured by the contact type technique such as the
hot-film probe, based on which the velocity and the turbulent kinetic energy could be
measured at single point. However, it is notable that for the flow with large void fraction
the hot film is easy to be burn out due to the inefficient cooling by gas phase. These
phenomena also limit the application of hot film to the turbulent bubbly flow
measurements. Nonetheless, it has provided us many useful databases to understand the
physics and develop the model of turbulent bubbly flow.
As for the measurement of the bubbles’ behaviors, generally the flow
visualization is a useful way which also requires high visibility of the fluid properties
and thus could only be adopted in few bubbles cases. At present, for the turbulent
bubbly flow with many bubbles, the bubbles’ behaviors are mainly measured by probes
such as optical-sensor probes and conductive-sensor probes, based on which the bubble
number, concentration and the bubble geometry could be measured at the single point.
Fig. 2.1 summarizes the applications of the current experimental techniques for the
turbulent bubbly flow measurement.
2.3 Experimental apparatus
The experimental apparatus for the upward turbulent bubbly flow generally
contains five systems including the liquid phase circulation loop, the gas phase
supplying system, the inlet two phase mixing or bubble generation system, the phase
separation system and the test section with the measurement systems. Fig. 2.2 shows a
simplified schematic diagram of the experimental apparatus.
For the single phase flow, when the flows are introduced to the ducts there exist
two stages including the developing region with the entrance effects and the developed
region far from the inlet and without the entrance effects. In order to establish and
validate the models, the universal databases are necessary, which requires the
39
experimental measurements to be carried out in the fully developed region or
developing region with accurate control of the inlet flow. However, unlike the single
phase flow where the fully developed flow status could be reached with the long enough
pipes, for the turbulent bubbly flow the following problems make it difficult to establish
the universal database for the model development.
(a) Inlet flow control: not only the flow velocities, but also the distributions of
the void fraction and the bubble diameters should be considered.
(b) Bubble coalescence and breakup: they could change the bubbles properties
but their measurements and mechanisms are still in an initial stage.
(c) Bubble expansion: the bubbles properties could be changed for the high and
long ducts so that the fully-developed turbulent bubbly flow is difficult to
reach.
In summary, for the upward turbulent bubbly flow, after it is introduced to the
ducts, the flow includes the developing of the liquid phase flow field and the dispersed
bubbles distributions or flow regimes which couples together and determines the final
flow characteristics. It is more complicated for the flow to reach the fully-developed
status for both the liquid phase flow field and the dispersed bubbles distributions.
During the establishment and validation the turbulent bubbly flow models, the
developing status is necessary to be examined with the carefully controlled and
measured inlet parameters such as the distributions of void fractions, the bubble
diameters, and the velocities of liquid phase and bubbles. Table 2.1 shows the locations
of the experimental measurements in the current used databases which are still very
limited to understand the flow developing process.
40
Table 2.1 Experimental databases in different duct geometries. B: Base; H: Height
Duct shape Experimental
database
Duct size
DH (mm)
Experimental
techniques
Measured locations
z/ DH Measurements
Small
Circular Pipe
Serizawa et al. [2.1-2.2] 60 Dual-senor Conductive
& Hot-film probes 10, 20, 30
α, vb, fb
W, wrms
Wang et al. [2.3] 57.15 Hot-film probe 35 α, W, wrms, vrms, <vw>
Liu et al. [2.4-2.5] 38
57.2
Dual-senor Conductive
& Hot-film probes
36
30, 60, 90, 120
α, vb, fb, Db,
W, wrms, vrms, <vw>
Hibiki et al. [2.6] 25.4
50.8
Dual-senor Conductive
& Hot-film probes
12, 65, 125
6, 30.3, 53.5
α, vb, Db, ai
W, wrms
Small
Triangle Duct Bertodano et al. [2.7]
B: 50
H: 100 Hot-film probe 73 α, W, wrms, vrms, <vw>
Large
Circular pipe
Shawkat et al. [2.8] 200 Dual-senor optical &
Hot-film probes 42
α, vb, fb, Db,
W, wrms, vrms, <vw>
Sun et al. [2.9] 100 Multi-senor Optical &
Hot-film probe 3, 18, 33 α, vb, Db, ai
Large
Square Duct Sun [2.10] 136×136
Multi-senor Optical &
Hot-film probe 4.5, 10.1, 16
α, vb, fb, Db, ai
W, wrms, vrms, <vw>
41
Fig. 2.1 (a) Experimental techniques for the flow measurements and (b) its
applications in the turbulent bubbly flow.
42
Fig. 2.2 Simplified schematic diagram of the experimental apparatus.
2.4 Experimental results
Firstly, the flow conditions of the collected databases were shown in Tables 2.2
based on the area-averaged superficial liquid velocity Jl and gas velocity Jg for different
geometries. Meanwhile, the range of the flow regime is also shown.
For simplicity, only several representative sets of experimental results were
shown here in Figs. 2.3–2.11 including the axial liquid phase velocities Wl, void fraction
α, the turbulent intensities k or turbulent components <uu> or <vv> and the measured
bubble diameter DB. For the other sets used in the dissertation, it could be found in the
corresponding references.
As shown from Figs. 2.3–2.11 and mentioned in Chapter 1, for the upward
43
turbulent bubbly flow in the small circular pipes, the pronounced wall peak of the void
fraction could be observed for low area-averaged void fraction flow conditions. With
the wall peak of void fraction, the average axial liquid velocity (Wl) and turbulence
profiles are more uniform in the core region. For the turbulent bubbly in the large
circular pipes, the void fraction profile tends to be flatter as compared to the turbulent
bubbly flow in the small circular pipes. The wall-peak of the void fraction could only be
observed for the cases with extremely low averaged void fraction. Generally, in the both
the small and large circular pipes, comparing to the upward single-phase flow the
steeper ∩-shape with flatter velocity profile in the core region could be observed with
the wall peak of the void fraction. However, in the large square duct, not only the
pronounced wall and corner peaks of the void fraction, but also the significant peaks of
Wl could be observed near the corner and wall which form like the typical M-shaped
velocity distribution between two parallel walls and two diagonal corners. Hence, here
the apparent M-shaped Wl profile is regarded as one of typical characteristics of the
upward turbulent bubbly flow in this large noncircular duct. The detailed mechanism
will be considered in the Chapter 5.
2.5 Summary
In this chapter, the experimental databases used in the present dissertation were
collected including the flow conditions and the representative experimental results. In
the following chapters, the numerical evaluation and further analysis will be carried out
based on these experimental databases. It is notable that, in this chapter, generally two
sets of the experimental results were shown for simplicity. The more detailed database
could be found in the corresponding references.
44
Table 2.2 Experimental flow conditions in different duct geometries. B: Bubbly flow; T: Transition regime.
Duct shape Experimental
database
Flow conditions
Flow regime
<Jl> <Jg> , χ (For Serizawa [2.1-2.2])
Small
Circular Pipe
Serizawa et al. [2.1-2.2]
0.74
0.88
1.03
0, 0.0119, 0.0240, 0.0358
0, 0.0099, 0.0198, 0.03
0, 0.0085, 0.0170, 0.0258, 0.0341
B
Wang et al. [2.3] 0.43 0, 0.1, 0.27, 0.4 B to T
Liu et al. [2.4-2.5]
0.376
0.535
0.753
0.0, 0.18
0.18
0, 0.18
B to T
Hibiki et al. [2.6] 0.491
0.986
0, 0.0275, 0.0556, 0.129, 0.190
0, 0.0473, 0.0113, 0.242, 0.321 B to T
Small Triangle
Duct Bertodano et al. [2.7] 1.0 0.1 B
Large
Circular Pipe
Shawkat et al. [2.8]
0.45
0.58
0.68
0.015, 0.03,0.05, 0.065, 0.085, 0.1
0.015, 0.03, 0.065, 0.085, 0.1, 0.18
0.015, 0.03, 0.065, 0.085, 0.1, 0.18
B to T
Sun et al. [2.9] 1.021 0.1 B
Large square
duct Sun et al. [2.10]
0.50 0.0, 0.068, 0.090, 0.135 B to T
0.75 0.0, 0.068,0.090, 0.135, 0.180 B to T
1.00 0.0, 0.090, 0.135, 0.180, 0.225 B to T
1.25 0.0, 0.135, 0.180, 0.225, 0.270 B to T
45
Fig. 2.3 Lateral distributions of axial liquid phase velocities Wl and void fraction α in
the small circular pipe with DH=60mm under (a) Jl =0.74 m/s and (b) Jl =1.03 m/s based
on Serizawa’s experimental database [2.1].
Fig. 2.4 Lateral distributions of axial turbulent intensities <ww> in the small circular
pipe with DH=60mm under (a) Jl =0.74m/s and (b) Jl =1.03m/s based on Serizawa’s
experimental database [2.1].
0 0.2 0.4 0.6 0.8 10.5
1
1.5
r/R
Wl
Jl=0.74m/s
X %
0.000
0.0119
0.024
0.0358
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0 0.2 0.4 0.6 0.8 1
0.5
1
1.5
r/R
Wl
J
l=1.03m/s
X %
0.000
0.0085
0.0170
0.0258
0.0341
0.0427
0.0511
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0 0.2 0.4 0.6 0.8 10
0.005
0.01
0.015
r/R
<w
w>
Jl=0.74m/s
X %
0.000
0.0119
0.024
0.0358
0 0.2 0.4 0.6 0.8 10
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
r/R
<w
w>
J
l=1.03m/sX %
0.000
0.0085
0.0170
0.0258
0.0341
0.0427
0.0511
(b) (a)
(a) (b)
46
Fig. 2.5 Lateral distributions of axial liquid phase velocities Wl and void fraction α in
the small circular pipe with DH=50.8mm under (a) Jl =0.491m/s and (b) Jl =0.986m/s
based on Hibiki’s experimental database [2.6].
Fig. 2.6 Lateral distributions of axial turbulent intensities <ww> in the small circular
pipe with DH=50.8mm under (a) Jl =0.491m/s and (b) Jl =0.986m/s based on Hibiki’s
experimental database [2.6].
0 0.2 0.4 0.6 0.8 10.4
0.6
0.8
1
r/R
Wl
Jl=0.491/s
Jg
0.0275
0.0556
0.129
0.190
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0 0.2 0.4 0.6 0.8 10.5
1
1.5
2
r/R
Wl
J
l=0.986/s
Jg
0.0473
0.113
0.242
0.321
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0 0.2 0.4 0.6 0.8 10
0.05
0.1
0.15
0.2
r/R
<w
w>
Jl=0.491/s
Jg
0.000
0.0275
0.0556
0.129
0.190
0 0.2 0.4 0.6 0.8 10
0.05
0.1
0.15
0.2
r/R
<w
w>
Jl=0.986/s
Jg
0.000
0.0473
0.113
0.242
0.321
(b) (a)
(b) (a)
47
Fig. 2.7 Lateral distributions of (a) axial turbulent kinetic energy <ww> and lateral turbulent kinetic energy <uu> and (b) void fraction α
in the small circular pipe with DH=60mm based on Liu’s experimental database [2.3-2.4].
0 0.2 0.4 0.6 0.8 10
0.01
0.02
0.03
<w
w>
0 0.2 0.4 0.6 0.8 1
0
0.01
0.02
r/R
<u
u>
Jl J
g
0.376 0.000
0.753 0.000
0.376 0.018
0.535 0.018
0.753 0.018
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
0.5
r/R
0.376 0.018
0.535 0.018
0.753 0.018
(a) (b) Jl Jg
48
Fig. 2.8 Lateral distributions of axial liquid phase velocities Wl and void fraction α in
the large circular pipe with DH=200mm under (a) Jl =0.45m/s and (b) Jl =0.68m/s based
on Shakwat’s experimental database [2.7].
Fig. 2.9 Lateral distributions of axial turbulent kinetic energy <ww> and lateral
turbulent kinetic energy <uu> in the large circular pipe with DH=200mm under (a) Jl
=0.45m/s and (b) Jl =0.68m/s based on Shakwat’s experimental database [2.7].
0 0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
Wl
Jl=0.45m/s
Jg
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
r/R
0.0000.050
0.0850.100
0 0.2 0.4 0.6 0.8 1
0.2
0.4
0.6
0.8
1
Wl
Jl=0.68m/s
Jg
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
r/R
0.0000.065
0.0850.100
0.180
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=0.45m/s
0 0.2 0.4 0.6 0.8 1
0
0.01
0.02
0.03
r/R
<u
u>
Jg
0.0000.050
0.0850.100
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=0.68m/s
0 0.2 0.4 0.6 0.8 1
0
0.01
0.02
0.03
r/R
<u
u>
Jg 0.000
0.0650.0850.100
0.180
(b) (a)
(b) (a)
49
Fig. 2.10 Lateral distributions of axial liquid phase velocities Wl and void fraction α in the large circular pipe with DH=136mm:
under Jl =0.75m/s (a) along the bisector line; (b) along the diagonal line; under Jl =1.00m/s (c) along the bisector line; (b) along the
diagonal line based on Sun’s experimental database [2.10].
0 0.2 0.4 0.6 0.8 10.5
1
1.5
Wl
Jl=0.75m/s Bisector
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
2x/DH
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10.5
1
1.5
Wl
Jl=0.75m/s Diagonal
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
2x/DH
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10.8
1
1.2
1.4
1.6
Wl
Jl=1.0m/s Bisector
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
2x/DH
0.0000.0900.135
0.1800.225
0 0.2 0.4 0.6 0.8 10.8
1
1.2
1.4
1.6
Wl
Jl=1.0m/s Diagonal
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
2x/DH
0.0000.0900.135
0.1800.225
(b) (a)
(d) (c)
Jg Jg
Jg
Jg
50
Fig. 2.11 Lateral distributions of axial turbulent kinetic energy <ww> and lateral turbulent kinetic energy <uu> in the large circular pipe
with DH=136mm under Jl =0.75m/s (a) along the bisector line; (b) along the diagonal line; under Jl =1.00m/s (c) along the bisector line;
(d) along the diagonal line based on Sun’s experimental database [2.10].
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=0.75m/s Bisector
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
2x/DH
<u
u>
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=0.75m/s Diagonal
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
2x/DH
<u
u>
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=1.0m/s Bisector
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
2x/DH
<u
u
0.0000.0900.135
0.1800.225
0 0.2 0.4 0.6 0.8 10
0.02
0.04
0.06
<w
w>
Jl=1.0m/s Diagonal
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
2x/DH
<u
u>
0.0000.0900.135
0.1800.225
(b) (a)
(d) (c)
Jg
Jg
Jg
Jg
51
Reference
[2.1] Serizawa, A., 1974, Fluid dynamics characteristics of two-phase flow, Doctor
Thesis, Kyoto University.
[2.2] Serizawa, A., Kataoka, I., Michiyoshi, I., 1975, Turbulence structure of air-water
bubbly flow—II. local properties, Int. J. Multiphase Flow 2, 235–246.
[2.3] Wang, S., Lee, S., Jones, O., Lahey, Jr R. T., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
[2.4] Liu, T. J., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
pipe—I. bubble void fraction, bubble velocity and bubble size distribution, Int. J.
Heat Mass Transfer 36, 1061–1072.
[2.5] Liu, T. J., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
pipe—I. liquid mean velocity and turbulence measurements, Int. J. Heat Mass
Transfer 36, 1049–1060.
[2.6] Hibiki, T., Ishii, M., Z. Xiao, 2001, Axial interfacial area transport of vertical
bubbly flows, Int. J. Heat Mass Transfer 44, 1869-1888.
[2.7] Bertodano, de M. L., 1992, Turbulent bubbly two-phase flow in a triangular duct.
Doctoral Thesis, Resselaer Polytechnic Institute.
[2.8] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2008, Bubble and liquid turbulence
characteristics of bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 34, 767-785.
[2.9] Sun, X., Smith, T. R., Kim, S., Ishii, M., Uhle, J., 2003, Interfacial structure of
air-water flow in a relatively large pipe. Exp. Fluids 34, 206-219.
[2.10] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
52
53
Chapter 3
Numerical evaluation of two-fluid model of turbulent
bubbly flow in large square duct
3.1 Introduction
As reviewed in Chapter 1, numerical efforts on turbulence modeling of bubbly
flow have been made based on an assumption of isotropic turbulence, such as k-ε model.
Moreover, the numerical evaluations of the previous models were mainly carried out
based on the experimental database in the small circular and noncircular ducts. However,
for the single-phase turbulent flow in the noncircular ducts, the secondary flow exists in
the cross-section due to the anisotropic turbulence, which has strong effects on the flow
characteristics such as distorting the local mean velocity distribution. Therefore, for the
turbulent bubbly flow in the noncircular ducts, the interaction between the secondary
flow and the bubbles could exist. In addition, the turbulent bubbly flow in the large duct
shows different flow characteristics from that in the small ducts. Thus, the models
mentioned above need to be evaluated for the turbulent bubbly flow in the large and
noncircular ducts. In this chapter, the numerical evaluation of the existed two-fluid
model will be carried out for the turbulent bubbly flow in the large square duct based on
the recently established database by Sun [3.1].
3.2 Numerical methodology
3.2.1 Problem description
The numerical evaluation here is carried out by mimicking previous simulation
54
of the turbulent bubbly flow in the circular pipe by changing the shape of the duct
geometry as shown in Fig. 3.1. Similar to that in [3.1], the duct size is chosen to be
0.136m×0.136m×3m. The bubble size DB is chosen according to the different
experimental conditions.
Fig. 3.1 Schematic diagram of the numerical problem.
3.2.2 Numerical methods
3.2.2.1 Governing equations
Considering the incompressible liquid and gas phase without phase change, the
governing equations of the turbulent bubbly flow based on Eulerian-Eulerian two-fluid
model [3.2] could be given as follows including the mass and momentum conservation.
/ 0gg gt V , (3.1)
55
/ 0ll lt V , (3.2)
Re
/ -
T
g g g g gg g g g g g g
g g g g g ig
t p
V V V V V
τ g M
, (3.3)
Re
/ -
T
l l l l ll l l l l l l
l l l l l il
t p
V V V V V
τ g M
, (3.4)
where the definitions of the variables are the same as that defined in Chapter 1. In
addition, the void fraction satisfies 1 gl ; the phase interaction through the
interfaces satisfies 0il ig M M ; the Reynolds stress term Reg g g τ in the gas phase
could be neglected due to lg .
3.2.2.2 Constitutive relations
The following constitutive relations were adopted for the interfacial and
turbulence terms to close the above equations:
For the interfacial term Mil, the phase interaction was considered by a drag force
MD [3.2], a virtual mass force MVm [3.3], a lift force ML [3.4], a turbulent dispersion
force MT [3.5], a wall force Mw [3.6], and a pressure force Mp[3.7].
il ig D VM L T W p M M M M M M M M . (3.5)
The following constitutive relations were adopted for each force:
3
- -4
g l g lD D l g
B
CD
M V V V V , (3.6)
-g l lL g L lC M V V V , (3.7)
Dg lVM g l VM Dt
C M V V , (3.8)
l
l
g
gtDTDT CC
-M , (3.9)
56
2
W -g lW g l wC M V V n , (3.10)
2
p -g lp g l iC M V V n , (3.11)
where CD, CL , CVM, CTD, CW , and Cp are the coefficients for each force and were
chosen as follows:
0.68724 72 8 Eomax min 1 0.15Re , ,
Re Re 3 Eo 4DC
;
min 0.288tanh 0.121Re , Eo For Eo 4
Eo For 4 Eo 10
0.29Eo For 10
m
L
f
C f
Eo
;
2/1VMC ,
wB
WyD
C05.001.0
,0max ; 1.0TDC ; 0.1pC
For the turbulence term τRel in Eq. (3.4), the η-ε model proposed by Lopez
Bertodano et al. [3.7] was adopted, which considers the turbulence in the liquid phase
τRel separately including the shear-induced turbulence τs and bubble induced turbulence
τB.
BS τττ Re . (3.12)
The shear-induced turbulence τs and bubble-induced turbulence τB are
respectively given by
SSl
T
llleffS kAVVτ 3
2, , (3.13)
BBlB kAτ 3
2 , (3.14)
where AS and AB are the anisotropy matrices. leff , is the effective eddy viscosity and
was chosen based on the k-ε model and Sato model [3.8] as
lgbglbSltbltleff uuDCkC /2
,, , (3.15)
where Cμl and Cμb are the model constants and chosen to 0.09 and 0.6 respectively.
57
To close Eq. (3.13) and (3.15), the shear-induced turbulent kinetic energy kS was
considered by the well-known k-ε model for the single-phase turbulent flow as
-Sll S l t S l
kk k P
t
V , (3.16)
2
1 2-tll l l
S S
PC C
t k k
V , (3.17)
where ,
T
l l leff lP
V V V ; ζε, Cε1 and Cε2 are model constants and are the
same as that in the single-phase k-ε model with ζε=1.3, Cε1=1.44, and Cε1=1.92,
respectively.
Considering the bubble-induced turbulent kinetic energy kB as a scalar, the
transport equations could be given as
-B l Tll B l t B Ba B
B
k Vk k k k
t D
V (3.18)
where VT is the terminal velocity of a single bubble and kBa is the asymptotic value of
the bubble-induced kinetic energy 21
2g lBa g VMk C V V and 2VMC
.
3.2.2.3 Boundary conditions
The boundary conditions especially at the wall are essential for the numerical
simulation and given as follows: In the inlet, the uniform boundary conditions were
adopted including uniform liquid and gas superficial velocities Jl and Jg and turbulence.
In the outlet, the natural flow conditions were adopted by giving the outlet pressure as
the atmosphere. At the wall, for the liquid phase, similar to the single-phase turbulent
flow, the wall function was used for the velocity and turbulence [3.9]. For the gas phase,
the void fraction at the wall was set to be 0Wallg indicating the bubble can’t enter
the wall.
58
3.2.2.4 Numerical procedures
In the present dissertation, the governing equations were numerically solved by
finite differential method, which could be briefly summarized as follows: Firstly, the
discretization of the above equations could be rearranged as
11 nn n n n n n
gg g g g g g A G V A , (3.19)
11 nn n n n n n
ll l l l l l A G V A , (3.20)
1 1n n nn n n n n n n
g g gVg g Vg Vg g Vgp A V H G V A V , (3.21)
1 1 n nn n n n n n n nl lVl l l Vl Vl l Vlp A V H G V A V , (3.22)
1n n n n n n n n
ks l S ks S ks l Sk k k A G A , (3.23)
1n n n n n n n n
s l S s S ks l S A G A , (3.24)
1n n n n n n n n
kB l B kB B kB l Bk k k A G A , (3.25)
where A, H and G are matrices related with the discretization of the time-derivative,
pressure term and the other terms including convection, diffusion, interfacial interaction,
turbulence and source terms.
Then, the projection method was used to solve the pressure filed p and update the
velocity field V as follows:
* /n nn n n ng gg Vg g Vg Vg V G V A V A , (3.26)
* /n nn n n n nl ll Vl l Vl Vl l V G V A V A , (3.27)
1 * 1 /
n n n n ng g Vg Vg gp V V H A , (3.28)
1 * 1 /
n n n n nl l Vl Vl lp V V H A , (3.29)
Substituting Eqs. (3.28) and (3.29) to the summation of Eqs. (3.19) and (3.20)
with the constraint of αg+αl=1 and rearranging, the pressure filed could be solved by
59
1 * *n n n n
l gp B C V D V , (3.30)
where B, C and D are matrices related with grid size.
With the new pressure field, the fields of velocities in liquid and gas phases, void
fraction and turbulence could be updated. Fig. 3.2 shows the flow chart which was used
to solve the governing equations.
3.3 Results and discussion
During the numerical calculation, the inlet flow rates and bubble size were set to
be similar to that in the experiments [3.1]. Figs 3.4 show the comparisons of the void
fraction, liquid phase velocities and turbulence between the experimental measurements
and predictions under Jl =0.75m/s and Jg =0.09m/s. For simplicity, the comparisons of
the other cases were not shown here since they showed the similar tendency. It is
obtained that the current k-ε model could predict the corner peak phenomenon of void
fraction and local liquid velocity. However, the predicted values of the peak void
fraction and velocity showed much smaller as compared with experiments, especially,
the turbulence fields.
The following factors are considered to be responsible for the above
insufficient prediction. (1) As compared with the experiments, the existing model
predicted the almost constant relative velocity Wr around 0.23m/s. However, in the
experiments, the relative velocity Wr was related with the void fraction and shows
smaller than the prediction. (2) Based on the k-ε model, the secondary flow effect was
neglected, which could affect the bubbles motion and push the bubbles to the corner. In
addition, unlike the circular pipe with a homogeneous boundary, the small bubbles
could be trapped in the corner region because of the non-homogeneous boundary. (3)
The model constants were mainly proposed and validated based on the experiments in
60
the small ducts. It might not be suitable in the large ducts however so far our current
understandings of the physical process therein are limited. Therefore, to improve the
model for the turbulent bubbly flow in the large noncircular ducts, firstly the
understanding of the physical process in the large noncircular ducts is urgently in need,
which is the target in the following four chapters.
3.4 Summary
In this Chapter, the numerical evaluations of the current two-fluid model of the
turbulent bubbly flow were carried out in the large square duct based on the recently
established database. It showed that the current model might not be suitable for the
prediction of the turbulent bubbly flow in the large noncircular ducts because of the
different physical processes therein as compared to that in the small circular pipes. In
order to improve the model therein, the further detailed analysis of the physical process
of the turbulent bubbly flow in the large noncircular ducts will be suggested such as the
liquid phase velocity peak near the wall, the significant corner peak of the void fraction,
and strong turbulence.
61
Fig. 3.2 Flow charts of numerical calculation.
62
Fig. 3.3 Comparisons of the experimental measurements and numerical model under Jl =0.75m/s and Jg =0.09m/s:
for the liquid phase velocity (a) along the bisector line and (b) along diagonal line; for the void fraction (c) along the bisector line and
(d) along diagonal line; for turbulence (e) along the bisector line and (f) along diagonal line.
0 0.2 0.4 0.6 0.8 10.7
0.8
0.9
1
1.1
2x/DH
Wl
Experimental data
Existing model
0 0.2 0.4 0.6 0.8 10
0.05
0.1
0.15
0.2
0.25
2x/DH
Experimental data
Existing model
0 0.2 0.4 0.6 0.8 10
0.005
0.01
0.015
0.02
2x/DH
t k
Experimental data
Existing model
0 0.2 0.4 0.6 0.8 10.7
0.8
0.9
1
1.1
2x/DH
Wl
Experimental data
Existing model
0 0.2 0.4 0.6 0.8 10
0.05
0.1
0.15
0.2
0.25
2x/DH
Experimental data
Existing model
0 0.2 0.4 0.6 0.8 10
0.005
0.01
0.015
0.02
2x/DH
t k
Experimental data
Existing model
(a)
(b)
(c)
(d)
(e)
(f)
63
Reference
[3.1] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
[3.2] Ishii, M., Hibiki, T., 2011, Thermo-fluid dynamics of two-phase flow. 2nd ed.
New York: Springer.
[3.3] Lahey Jr. R.T., 1990, The analysis of phase separation and phase distribution
phenomena using two-fluid models, Nucl. Eng. Des. 122, 17-40.
[3.4] Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002, Transverse migration of
single bubbles in simple shear flows, Chem. Eng. Sci. 57, 1849-1858.
[3.5] Burns, D., Frank, T., Hamill, I., Shi, J. M., 2004, The Favre averaged drag
model for turbulent dispersion in Eulerian multi-phase flow, Proc. 5th
International Conference on Multiphase Flow, Yokohama, Japan, 392.
[3.6] Antal, S.P., Lahey, Jr R. T., Flaherty, J. E., 1991, Analysis of phase distribution
in fully developed laminar bubbly two-phase flow, Int. Multiphase Flow 13,
309-326.
[3.7] Bertodano, Lopez M., Lahey, Jr. R.T., Jones, O. C., 1994, Phase distribution in
bubbly two-phase flow in vertical ducts, Int. J. Multiphase Flow 20, 805-818.
[3.8] Sato, Y., Sadatomi, M., 1981, Momentum and heat transfer in two-phase bubble
flow—I. Theory, Int. J. Multiphase Flow 7, 167-177.
[3.9] Pope, S. B., 2000, Turbulent flows, Cambridge University Press.
64
65
Chapter 4
Developing process of turbulent bubbly flow
4.1 Introduction
Several researches have discussed the developing of the turbulent bubbly flow
but mainly focused on the general flow pattern transition from the bubbly flow to the
slug flow [4.1-4.3]. Few researches were focused on the detailed developing process
based on the bubble motion. Consequently, there does not exist any detailed model
about the formation of the bubble layer, the velocity and the turbulence distribution
during the developing process. Similar to the single-phase turbulent flow, in order to
develop the developing models for the turbulent bubbly flow, the experimental
databases are essentially required with the well-controlled inlet flow conditions
including not only velocity distributions but also the bubble distribution. However, for
the circular pipes the experimental databases could not satisfy the requirements which
were either with complicated inlet flow conditions [4.4-4.5] or measured far from the
inlet [4.6]. In Sun’s experiments [4.7], the experimental measurements were carried out
at different locations during the flow developing with well-controlled inlet flow
conditions, which provides us the possibility to consider the developing process. In this
chapter, the developing process in the turbulent bubbly flow will be discussed based on
the database in [4.7].
4.2 Experimental observations
Firstly, based on Sun [4.7] Figs. 4.1-4.4 show the experimental results of the
66
liquid phase velocities and the corresponding void fractions at different locations for the
different flow conditions summarized in Table 4.1. From the inlet to the measured
location, generally two stages could be observed during the developing including the
formation of the bubble layer near the wall and the developing the liquid phase with this
bubble layer the image of which was shown in Fig. 4.5. Moreover, the developing
processes for the small void and large void fraction cases in the transition region
showed quite different characteristics. For the cases with large void fraction, after the
formation of the bubble layer, the further rearrangement of the bubbles could be
observed due to the coalescence of the bubbles occur near the wall, which made the
bubble lateral migration from the wall to the center. This process dominates the second
stage of the developing process and was more complicated. In this chapter, we will
discuss the developing processes during these two stages especially for the small void
fraction. The following topics will be focused on. In the first stage, the formation of the
bubble layer will be discussed by considering how long it needs to form the developed
bubble layer. In the second stage, the length required to lift up the liquid phase velocity
will be discussed.
Table 4.1 Experimental conditions for the developing process.
Jl (m/s) 0.5 0.75 1.0 1.25
Jg (m/s) 0.09 0.135 0.09 0.135 0.18 0.09 0.135 0.18 0.135 0.18 0.225
67
Fig. 4.1 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5 DH, 10 DH and 16DH for Jl=0.5m/s:
with Jg=0.09m/s along (a) the bisector line; (b) the diagonal line; with Jg=0.135m/s along (c) the bisector line; (d) the diagonal line [4.7].
0
0.5
1
Wl
Jl=0.5m/s J
g=0.09m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
Wl
Jl=0.5m/s J
g=0.09m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
Wl
Jl=0.5m/s J
g=0.135m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
Wl
Jl=0.5m/s J
g=0.135m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
(a) (b)
(c) (d)
68
Fig. 4.2 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5 DH, 10 DH and 16DH for Jl=0.75m/s:
with Jg=0.09m/s along (a) the bisector line; (b) the diagonal line; with Jg=0.135m/s along (c) the bisector line; (d) the diagonal line;
with Jg=0.18m/s along (e) the bisector line; (f) the diagonal line [4.7].
0
0.5
1W
l
Jl=0.75m/s J
g=0.09m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
Wl
Jl=0.75m/s J
g=0.135m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl
Jl=0.75m/s J
g=0.18m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl
Jl=0.75m/s J
g=0.09m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl
Jl=0.75m/s J
g=0.135m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl
Jl=0.75m/s J
g=0.18m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
(a)
(b)
(c)
(d)
(e)
(f)
69
Fig. 4.3 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5 DH, 10 DH and 16DH for Jl=1.0 m/s:
With Jg=0.09m/s along (a) the bisector line; (b) the diagonal line; with Jg=0.135m/s along (c) the bisector line; (d) the diagonal line;
with Jg=0.18m/s along (e) the bisector line; (f) the diagonal line [4.7].
0
0.5
1
1.5
Wl
Jl=1.0m/s J
g=0.09m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl J
l=1.0m/s J
g=0.135m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
1
2
Wl
Jl=1.0m/s J
g=0.18m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl
Jl=1.0m/s J
g=0.09m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
0
0.5
1
1.5
Wl J
l=1.0m/s J
g=0.135m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
0
1
2
Wl
Jl=1.0m/s J
g=0.18m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
(a)
(b)
(c)
(d)
(e)
(f)
70
Fig. 4.4 Liquid phase velocities Wl and the corresponding void fractions α at z=4.5 DH, 10 DH and 16DH for Jl=1.25m/s:
with Jg=0.135m/s along (a) the bisector line; (b) the diagonal line; with Jg=0.18m/s along (c) the bisector line; (d) the diagonal line;
with Jg=0.225m/s along (e) the bisector line; (f) the diagonal line [4.7].
1
1.5
2W
l
Jl=1.25m/s J
g=0.135m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
1
1.5
2
Wl
Jl=1.25m/s J
g=0.18m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
1
1.5
2
Wl
Jl=1.25m/s J
g=0.225m/s Bisector
0 20 40 60
0
0.2
0.4
Distance from the wall x mm
z=4.5DH
z=10DH
z=16DH
1
1.5
2
Wl
Jl=1.25m/s J
g=0.135m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
1
1.5
2
Wl
Jl=1.25m/s J
g=0.18m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
1
1.5
2
Wl
Jl=1.25m/s J
g=0.225m/s Diagonal
0 20 40 60 80 100
0
0.2
0.4
Distance from the corner x mm
z=4.5DH
z=10DH
z=16DH
(a)
(b)
(c)
(d)
(e)
(f)
71
Fig. 4.5 Schematic diagram of the developing process of the upward turbulent bubbly
flow in the large square duct.
4.3 Modeling of developing process
4.3.1 Formation of bubble layer in the first stage
In order to determine the length required for the formation of the bubble layer
LδB, the following two questions are necessary to be considered, i.e., why the lateral
migration of the bubbles occurs and when the stable bubble layer could be formed. As
mentioned in Chapter 1, the bubble lateral accumulation in the duct flow results from
the lift force because of the liquid phase velocity gradient in the lateral direction which
is dominated by the viscosity and the wall effect. Therefore, the process of the bubble
layer formation could be modeled as follows. Given the bubbly flow with perfect inlet
boundary conditions which are the uniform distributions of the bubbles liquid phase
velocity and turbulence, two processes coupled together during the flow developing
process including the developing of the liquid boundary layer and the bubble layer.
Then, when the bubble layer could be formed? The lateral force balance exerted on the
bubbles as shown in Fig. 4.6 dominates the bubbles lateral migration and the formation
of the bubble layer. In addition to the wall lubrication force [4.8], the turbulence
developing near the wall due to the shear and bubbles could induce the lateral turbulent
dispersion force [4.9] toward the center. Moreover, the coalescence of the bubbles due
72
to accumulation of the bubbles near the wall will also occur, which increased the bubble
size and decreased the lift force. When the lateral forces reached the balanced status, the
bubble layer could be formed. Herein, the flow with the small void fraction without the
bubble coalescence effect will be only considered. Therefore, the developing process of
the bubble layer is dominated by the developing of the liquid phase boundary layer
including the liquid phase velocity which generated the lateral lift force and the
turbulence near the wall which generated the turbulent dispersion force balanced with
the lift force. The length required to form the bubble layer LδB then depends on the
developing process of the liquid phase boundary layer.
Fig. 4.6 Schematic diagram of the formation of the bubble layer due to the lift force
exerted on the bubbles and the developing of the boundary layer in the liquid phase.
The following method is used to determine the developing length of the bubble
layer LδB. Firstly, the following assumptions are considered: (1) The uniform velocity
and bubble distribution in the inlet; (2) The little bubble effect on the liquid flow
developing during the lateral accumulation of the bubbles; (3) Neglecting the turbulent
73
dispersion force; (4) During the formation of boundary layer, only the neighbor bubble
could enter this layer immediately and other bubbles will also be mixed immediately.
With the above assumptions, the formation of the bubble layer with the thickness δB at
least requires the existence of the liquid phase boundary layer with the thickness δl = δB,
i.e., the length from the inlet for the bubble layer formation LB could be reflected by the
length from the inlet for liquid-phase boundary layer formation with the thickness Lδl.
Fig. 4.7 Comparison between the thickness of the boundary layer δl and the bubble layer
δB at the measurement location z=4.5DH.
Then, the bubble effect on the flow developing and the turbulent dispersion
force effect will be considered. As compared to the single-phase flow, the liquid phase
velocity in the boundary layer will be lifted up by the bubble accumulation and
buoyancy effect therein, which will lead to reduction of the lift force. The turbulent
dispersion force exerted on the bubbles will resist the bubbles moving into the boundary
layer. Therefore, the thickness of the bubble layer δB will be smaller than that of the
boundary layer δl, or the developing length LB will be shortened by these two effects,
0 0.005 0.01 0.015 0.02 0.025 0.030
0.5
1
1.5
2
Turbulent boundary layer thickness m
Dis
tan
ce f
rom
th
e u
nif
orm
in
let
z m
/z = 0.37Rez
-1/5
Wm
=0.50 m/s
Wm
=0.75 m/s
Wm
=1.00 m/s
Wm
=1.25 m/s
Thickness of bubble layer
z=4.5DH
z=3.3DH
z=2.8DH
Bubble layer formed
74
i.e., LB > Lδl. Here, the developing process of the liquid phase boundary layer will be
considered based on the single-phase turbulent boundary layer theory [4.10]:
1/5/ 0.371Rel zz , where Re /z Vz . Fig. 4.7 shows the thickness of the boundary
layer δl during the developing and the bubble layer δB at the measurement location
z=4.5DH. It could be observed that based on the above assumptions δl > δB at the
measurement location z=4.5DH, indicating the bubble layer could be already formed
before entering the measurement location.
4.3.2 Liquid phase velocity lifting in the second stage
Due to the formation of the bubble layer near the wall which made the fluid
lighter than that in the core region, the liquid phase velocity near the wall will be lifted
up by the bubbles. Fig. 4.8 shows the schematic diagram of liquid phase velocity lifting
up in the second stages as shown in Fig. 4.5. In the part, the length required to lift up the
liquid phase velocity Lp will be discussed based on the dimensional analysis as follows.
Fig. 4.8 Schematic diagram of liquid phase velocity lifting up in the second stage.
After forming the bubble layer near the wall, the liquid phase therein will be
accelerated due to the buoyancy effect. Here, considering the force balance on the liquid
75
phase, the acceleration rate al could be estimated as
1/gal . (4.1)
For different mean liquid phase velocity W, the time tl required to lift the velocity near
the wall from h1×W to h2×W could be estimated as
2 1 1 /lt h h W g . (4.2)
Then the length Lp required to lift the velocity near the wall from 0 to h×W could be
estimated as considering the acceleration
2 2 2
2 1 1 / 2pL h h W g . (4.3)
The dimensionless parameter A is defined as follows which shows the
developing status of the liquid phase velocity and ratio between the measured location
zp and Lp:
2 2 2
2 12 / 1p
p
p
zA z g h h W
L
, (4.4)
Fig. 4.7 shows the ratio A based on the experimental results corresponding to the
cases as shown in Table 4.1. For these cases, the measurements were carried out for
three locations zp =4.5DH, 10DH and 16DH. There might exist a critical value AC,
according to which, the velocity developing status could be judged by
For not lifted up
For lifted up
C
C
A A
A A
. (4.5)
According to the experimental results shown in Fig. 4.2, it could be obtained
that for the cases with Jl = 0.75m/s and Jl = 0.09m/s, the liquid phase velocities near the
wall shows still flat at 4.5DH and then were lifted up at z=10DH and stable until 16DH.
Thus, the dimensionless parameter A for this case at z=10DH is chosen as a criterion AC.
76
Correspondingly, the following conjectures for the other cases could be obtained: For
the cases with Jl = 0.5m/s, the liquid phase velocities could be lifted up before 4.5DH.
For the cases with Jl = 1.0m/s, the liquid phase velocities could be lifted up around or
after 10DH. For example, it is around 10 DH for Jg = 0.18m/s, 16 DH for Jg = 0.135m/s,
but more than 16 DH for Jg = 0.09m/s. For the cases with Jl = 1.25m/s, except Jg =
0.225m/s, more than 16 DH is required to lift up the liquid phase velocities near the wall.
By comparing with Figs. 4.1 -4.4 the above conjectures could be well confirmed, which
indicates the dimensionless parameter A could be a proper criterion to judge the flow
developing status.
Fig. 4.9 Comparison of dimensionless parameter 2/ 1pA L g W for different
superficial liquid and gas phase velocities Jl and Jg and measurement locations.
0 0.5 1 1.5 2 2.50
5
10
15
20
Jl
A
z=4.5DH
z=10DH
z=16DH
Ac=2.8
0.5 1 1.50
2
4
6
<Jl>
A
Jl
77
4.4 Summary and Discussion
In Section 4.2-4.3, the developing processes of the turbulent bubbly flow were
analyzed based on Sun’s experimental results [4.7]. Two stages were observed during
the flow developing which are forming the bubble layer and the velocity developing
based on the formed bubble layer. At the first stage of the bubble layer formation, the
length required to form the bubble layer was analyzed, which was found to be less than
that required to develop the liquid phase boundary layer with the similar thickness. At
the second stage of the velocity developing, the length required to lift up the liquid
phase velocity near the wall was analyzed, based on which the criterion to judge the
flow developing status was suggested.
However, so far the lack of the detailed experimental databases during the flow
developing process strongly limited us to verify the above analysis and to further
understand and model of this process. Many interesting questions and modeling are still
unable to be considered including:
(1) For the turbulent bubbly flow the thickness of the bubble layer δB was
observed to around 2δB/DH ≈0.2. And why the bubble layer was formed with
thickness? During the formation of the bubble layer, how do the distributions
of the phase, liquid velocity and turbulence evolve?
(2) At second stage where the velocity develops with bubble layer, how much
does it need to develop the velocity such as lift up the local liquid velocity?
After lifting up the liquid phase velocity near the wall, how are the bubbles
and liquid phase velocity rearranged?
Therefore, in order to understand the developing process of the turbulent
bubbly flow, the boundary layer flow type with careful design and detailed
measurements are suggested.
78
Reference
[4-1] Kaichiro, M., Ishii, M., 1984, Flow regime transition criteria for upward
two-phase flow in vertical tubes, Int. J. Heat Mass Transfer 27, 723–737.
[4-2] Schlegel, J.P., Sawant, P., Paranjape, S., Ozar, B., Hibiki, T., Ishii, M., 2009,
Void fraction and flow regime in adiabatic upward two-phase flow in large
diameter vertical pipes, Nucl. Eng. Des. 239, 2864–2874.
[4-3] Ohnuki, A., Akimoto, H., 2000, Experimental study on transition of flow pattern
and phase distribution in upward air–water two-phase flow along a large vertical
pipe, Int. J. Multiphase Flow. 26, 367–386.
[4-4] Serizawa, A., 1974, Fluid dynamics characteristics of two-phase flow, Doctor
Thesis, Kyoto University.
[4-5] Wang, S., Lee, S., Jones, O., Lahey, Jr R. T., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
[4-6] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2008, Bubble and liquid turbulence
characteristics of bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 34, 767-785.
[4-7] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
[4-8] Antal, S.P., Lahey, Jr R. T., Flaherty, J. E., 1991, Analysis of phase distribution
in fully developed laminar bubbly two-phase flow, Int. Multiphase Flow 13,
309-326.
[4-9] Lahey, Jr R. T., , Bertodano, de M. L., Jones, O. C., 1993, Phase distribution in
complex geometry conduits, Nucl. Eng. Des. 141,177-201.
[4-10] Schlichting, H., Gersten, K., 2008, Boundary layer theory, 8th
ed., Springer.
79
Chapter 5
Axial liquid-phase velocity of turbulent bubbly flow
5.1 Introduction
As shown in Fig. 2.9, for the turbulent bubbly flow in the large square duct [5.1],
not only the pronounced wall and corner peaks of void fraction, but also the significant
peaks of liquid phase velocities were observed near the corner and wall which form like
the typical M-shaped velocity distribution as shown in Fig. 5.1(a) between two parallel
walls, especially two diagonal corners. Generally, for the upward turbulent bubbly flow
in the small and large circular pipe, the steeper ∩-shape with flatter velocity profile in
the core region as compared with the upward single-phase flow was observed [5.2-5.6]
even with the significant wall peak of the void fraction. Therefore, the apparent
M-shaped Wl profile could be regarded as one of typical characteristics of the upward
turbulent bubbly flow in this large square duct where both the scale and the
non-homogeneous geometrical boundary effects are important. In this chapter, we firstly
discussed the mechanisms for formation of typical M-shaped Wl profile observed in the
previous upward turbulent bubbly flow experiments [5.1]. Here, three topics about the
formation of this typical velocity profile will be firstly discussed including the index of
flow type, the conditions for the formation of the M-shaped Wl profile and the peak
location of Wl. Based on that, the duct geometry effects on the void fraction profile and
wall shear stress will be considered. Then, the flow characteristics will be discussed in
two layers based on the normalization of obtained axial liquid phase velocity. Finally,
the numerical validation of the axial liquid phase velocity will be carried out.
80
Fig. 5.1 Schematic diagram of the type of the velocity profile of the upward flow:
(a) M-shape; (b) ∩-shape; (c) Cuspid-shape; (d) W-shape
5.2 Index of flow type
In this part, the index of the flow type as shown in Fig. 5.1 will be focused on in
order to consider the formation conditions of M-shaped Wl profile. Generally the
formation of different velocity profile in the upward turbulent bubbly flow could be
attributed to the counteraction between the wall and bubble effect which exert retarding
and pushing effect on the liquid flow, respectively.
To begin with, the wall effects on upward flow driven by pressure drop were
considered. Generally, the wall retards the main flow because of the nonslip condition at
the wall and fluid viscosity. For the single-phase upward flow with the constant density
the wall retarding effect is exerted on the whole area, so the ∩-shaped velocity profile
shown in Fig. 5.2(b) could be formed. However, for the upward turbulent bubbly flow,
the bubbles lateral migration due to the lift force result in the non-uniform density of the
fluid ρf =αlρl +αgρg. Hereafter, the symbols α and ρ are the void fraction and density,
and the subscripts ―l‖ and ―g‖ indicate the liquid and gas phase, respectively. The area
with wall retarding effect could be modified by the non-uniform density distribution.
To show the coupling interaction between the wall and non-uniform density
effect, the local net driving force F was evaluated. Considering the mixture of the gas
and liquid phases, the averaged local net driving force F directed to the upward in the
81
cross section could be calculated by
ggdzdPF ggll / , (5.1)
where dp/dz is the mean pressure gradient of the mixture and the main driving force
coming from the pump and compressor; g is gravity acceleration: -9.81m2/s. The gas
phase could be neglected since ρg << ρl. Here assuming fully-developed flow status and
neglecting the flow acceleration and secondary flow, the net driving force F is balanced
with the local net viscous force dηv/ds (ηv: local viscous force and s: coordinate direction,
x or y) from the neighbor fluid elements due to the molecular and turbulent eddy
viscosities.
Next, we consider the indications of physical process between the wall and
center by the sign of local net driving force F. If locally F > 0, it indicates the pressure
gradient is able to overcome the local gravity, and the flow could be pushed upward
driven by the pressure gradient. The local net viscous force from the neighbor fluid
element is downward and retards the local flow as a flow resistance. But if locally F < 0,
it indicates the pressure gradient is unable to overcome the local gravity. The fluids tend
to fall downward and keep the flow upward because the local net viscous force is
upward which pulls the flow as the additional driving force. In this case, the flow is
driven by the viscous shear force.
For the fully-developed turbulent flow, we assume the pressure gradient dp/dz to
be uniform among the cross sections. With the above indication, for the upward
single-phase flow driven by the pressure gradient, the net driving force F is uniform and
positive. Due to the wall, the local net viscous force always acts as the flow resistance
and the ∩-shaped velocity profile is formed. However, for the upward turbulent bubbly
flow, the net driving force F is non-uniform due to the non-uniform density among the
cross sections. According to the sign of F, the formation of different shape velocity
82
profile could be implied.
In [5.1], the mean pressure gradient dp/dz and αg distributions were measured
simultaneously and assuming dp/dz to be uniform among cross sections, the lateral
distributions of F were calculated and shown along the bisector and the diagonal lines,
respectively in Figs. 5.2-5.3. Table 5.1 shows the experimental data of the mean
pressure gradient dp/dz for different flow conditions at z=16DH based on the pressure
and height difference between z=15DH and z=19DH measured by the differential
pressure transmitters. Besides, the void fraction distribution at z=10.1DH was used since
the flow at z=16DH was developed based on void fraction distribution before entering
this part such as z=10.1DH. Besides, the void fraction distribution has already reached
the developed status at z=4.5DH as discussed in Section 4.2, so the choice of void
fraction distribution at different locations have no effect on the following analysis.
Table 5.1 Experimental data of dp/dz (kPa/m).
Jg (m/s)
Jl (m/s)
0.090 0.135 0.180 0.225 0.315
0.75 8.90 8.50 8.12 7.86 -
1.00 9.10 8.78 8.48 8.19 7.98
As shown in the above figures, the obvious sign changing of F at point xp near
the wall and corner could be observed for the turbulent bubbly flow in the large square
duct. Moreover, completely different behaviors were observed for the small,
intermediate and large void fraction cases. For the small void fraction case, the local net
driving force F is positive between xp and the wall, but negative between xp and the duct
center, indicating that the local net viscous force behaves as flow resistance between xp
83
and wall but driving force between the point xp and the duct center. Because of the wall
retarding effect, for the small void fraction cases the viscous force drags the flow near
the wall. After xp, the flow is driven by the local viscous force due to the high velocity
near the wall, indicating the wall retarding effect disappears or is blocked by the
bubbles accumulated between the wall and xp. It is notable that when F has a sign
changing point, the cuspid-shaped velocity profile shown in Fig. 5.1(c) also implying
the role of local net viscous force changes from the drag force to the driving force.
However, since there is no driving force in the duct center, this flow type could not be
formed. Therefore, the flow in this area is pulled upward by the flow near the wall from
falling down and driven by the shear. Therefore, with this distribution of F, it implies
the liquid phase velocity peak could be observed around the point xp where the sign
change of F occurs, and the M-shaped Wl profile. The following flow types were also
inferred for the upward turbulent bubbly flow in the large square duct as shown in Fig.
5.4. Due to the bubbles accumulation near the wall, the lighter mixture layer was
formed and pushed upward faster by the pressure gradient. In the core region, the liquid
is heavier and the flow is pulled upward by the faster layer near the wall. In other words,
two layers with the different flow types are formed between the wall to the duct center
including the upward pressure driven flow near the wall, and the viscous shear driven
flow in the core region. Wall retarding effect was confined in the lighter mixture layer or
blocked from transferring into bulk region. Moreover, if xp gets closer to the wall, the
wall effect is confined in a smaller area and blocked more.
To verify the above viewpoints, the corresponding lateral distributions of liquid
phase velocity were shown in Fig. 5.2(b) and Fig. 5.3(b) along the bisector line and
diagonal line for the bubbly flow, respectively. It is confirmed that once the local net
driving force F has an obvious sign changing near the wall, the lateral distribution of the
84
axial liquid-phase velocity Wl behaves as M-shaped. Besides, the location of the
velocity peak is observed to be lied close to xp according to the current experimental
data. Therefore it is suitable to choose F as an index of the flow type.
For the cases with the larger void fraction, since the bubbles coalescence occurs
and the bubbles were migrated to the core region, the net driving force F is negative
between xp and the wall, but positive between xp and the duct center. It indicates the net
viscous force acts as the driving force upward between the point xp and the wall, but the
flow resistance between xp and the duct center. Similarly, two layers with different flow
types as shown in Fig. 5.5 were also formed between the wall to the duct center but with
the viscous shear driven flow near the wall pulled by the faster flow in the duct center.
With further increasing the void fraction, more bubbles moved to the core region which
led to the heavier fluid near the wall. Once the viscous shear force is not strong enough
to overcome the local gravity, the W-shaped velocity profile with the downward flow
near the wall as shown in Fig. 5.1(d) is possible to be observed. However, so far the
current experimental technique (the X type hot-film anemometry used here) can’t
measure the liquid phase velocities under these flow conditions due to the limitations of
the hot-film anemometry measurement due to the high void fraction. Therefore,
currently it is lack of the experimental data of liquid velocities to validate the W-shaped
velocity profile.
In addition, for the intermediate void fraction cases in the transition regime such
as Jl=0.75m/s and Jg=0.18m/s and Jl = 1.0m/s, Jg = 0.27m/s, it is notable that along the
diagonal line, double peaks of F were observed as shown in Fig.5.3 (a) and 5.3(c) due to
the double void peak. For these cases, the peak locations of the liquid velocities tend to
migrate to the duct center especially along the diagonal line. Along the bisector line, the
peak locations of liquid velocities deviate more from xp due to the following reasons. As
85
we have confirmed, for the bubbly flow in this large square duct under certain condition,
peaks of void fraction and liquid phase velocity could be observed near the wall center
and corner with more apparent peak near the corner. For the small void fraction cases,
the locations of F=0 along the diagonal Xp1 and bisector Xp lines are close to each other.
However, for the cases in the transition regime such as Jl = 0.75m/s, Jg = 0.18m/s and at
Jl = 1.0m/s, Jg = 0.225m/s, Xp1 and Xp tends to be farther to each other, with Xp1 closer to
the duct center as shown in Fig. 5.6. And thus the location of the velocity peak along the
bisector line might be affected by that along the diagonal line and gets closer to the duct
center.
86
Fig. 5.2 Distributions of the local net driving force F for (a) Jl=0.75m/s and (c) Jl=1.0m/s and corresponding liquid velocities Wl for (b)
Jl=0.75m/s and (d) Jl=1.0m/s along the bisector line.
0 10 20 30 40 50 60 70-1500
-1000
-500
0
500
1000
1500
Distance from the wall x (mm)
Lo
cal
net
dri
vin
g f
orc
e F
0.09
0.135
0.18
0.27xp
(a)
Jl=0.75 m/s
Jg (m/s)
0 10 20 30 40 50 60 700.5
0.6
0.7
0.8
0.9
1
1.1
1.2
Distance from the wall x (mm)
Liq
uid
vel
oci
ty W
l(m/s
)
0.0
0.09
0.135
0.18xp
(b)
Jg (m/s)
Jl=0.75 m/s
0 10 20 30 40 50 60 70-1500
-1000
-500
0
500
1000
1500
Distance from the wall x (mm)
Lo
cal
net
dri
vin
g f
orc
e F
0.09
0.135
0.18
0.225
0.315xp
(c)
Jl=1.0 m/s
Jg (m/s)
0 10 20 30 40 50 60 700.7
0.8
0.9
1
1.1
1.2
1.3
1.4
Distance from the wall x (mm)
Liq
uid
vel
oci
ty W
l (m
/s)
0.0
0.09
0.135
0.18
0.225xp
(d)
Jl=1.0 m/s
Jg (m/s)
high void fraction
high void fraction
low void fraction
low void fraction
87
Fig. 5.3 Distributions of the local net driving force F for (a) Jl=0.75m/s and (c) Jl=1.0m/s and corresponding liquid velocities Wl for (b)
Jl=0.75m/s and (d) Jl=1.0m/s along the diagonal line.
0 10 20 30 40 50 60 70-2000
-1500
-1000
-500
0
500
1000
1500
Distance from the wall x (mm)
Lo
cal
net
dri
vin
g f
orc
e F
0.09
0.135
0.18
0.27
(a)
xp1
xp2
Jg (m/s)
Jl=0.75 m/s
0 10 20 30 40 50 60 700.4
0.6
0.8
1
1.2
1.4
Distance from the wall x (mm)
Liq
uid
vel
oci
ty W
l(m/s
)
0.0
0.09
0.135
0.18xp1
xp2
(b)
Jl=0.75 m/s
Jg (m/s)
0 10 20 30 40 50 60 70-1500
-1000
-500
0
500
1000
1500
2000
Distance from the wall x (mm)
Lo
cal
net
dri
vin
g f
orc
e F
0.09
0.135
0.18
0.225
0.315
xp1 x
p2
(c)
Jl=1.0 m/s
Jg (m/s)
0 10 20 30 40 50 60 70
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
Distance from the wall x (mm)
Liq
uid
vel
oci
ty W
l(m/s
)
0.0
0.09
0.135
0.18
0.225xp1 x
p2
(d)
Jl=1.0m/s
Jg (m/s)
high void fraction
low void fraction
low void fraction
high void fraction
88
Fig. 5.4 Schematic diagram of flow image for the small void fraction: (a) Bubble lateral
distribution; (b) Inside the bubble layer near wall; (c) Outside the bubble layer.
Fig. 5.5 Schematic diagram of flow image for the large void fraction: (a) Bubble
lateral distribution; (b) Flow near the wall; (c) Flow in the core region.
Fig. 5.6 Schematic diagram of velocity peak location.
89
5.3 Conditions for M-shaped Wl profile
As mentioned previously, except in the large square duct, in the existed database
of upward turbulent bubbly flow, there were very few observations of such significant
liquid-phase velocity peak near the wall region except that slightly peak or slightly shift
of velocity peak from the pipe center which is termed as ―chimney effects‖[5.3]. In the
small pipe, generally the steeper ―∩‖ shape or slightly shift of velocity peak was
observed even with significant wall peak of void fraction. Following the analysis in
section 4.3.1, in this section the conditions required for the formation of M-shaped Wl
profile will be discussed.
Firstly, based on the fully-developed flow the time averaged force balance
globally, we have
HggdzdP wgmglml / , (5.2)
where αlm=1-αgm and αgm are the mean volume fraction of the liquid and gas phases,
respectively. ηw is the wall shear stress per unit area exerted on the flow and is negative
if the wall drags the flow, and is also important to determine the internal flow structures.
H is the geometry dimension depending on the different geometry, and given for several
geometry in Table 5.2.
Table 5.2 Geometry dimension H to determine the wall shear stress
Duct
Geometry
Circular pipe
DH: pipe diameter
Channel
DH: Channel width
Square duct
DH: Duct width
Triangle
DH: Side length
H DH /4 DH /2 DH /4 DH /12
Substituting Eq. (5.2) to Eq. (5.1), the local net driving force F could be
rearranged as follows:
HgF wggmGl / . (5.3)
90
By setting Eq. (5.3) equals to zero, the sign changing point of F satisfies
gHGlwgmg 0 , (5.4)
where αg0 is the void fraction at the position where F changes the sign. Since αg
0 > 0, it
also requires
gHGlwgm . (5.5)
According to the previous analysis in Section 4.3.1, the M-shaped Wl profile is
possible to be formed when
pxx ,0 , 0
gg and cp xxx , , 0
gg , (5.6)
where xp is the distance between the location of the sign changing point of F and the
wall x=0. Besides, it is easier to observe M-shaped velocity profile if xp is closer to be
zero, since the wall effect was confined in a narrower region near the wall.
Eqs. (5.5) and (5.6) indicate that under the similar wall shear stress ηw/H, the
cases with larger αgm and steeper profile of the void fraction is easier to form M-shaped
velocity profile. Under the similar void fraction distribution, the cases with the smaller
wall shear stress ηw/H are easier to form the M-shaped velocity profile. The final flow
type is determined by the lateral void fraction distribution and the wall drag resistance.
Under the same flow conditions, duct geometry plays a dominated effect on determining
these two factors.
5.4 Duct geometry effect
To answer the reason why the typical M-shaped velocity profile is observed in
the large square duct rather than the small circular and non-circular ducts and also the
large circular pipe, the duct geometry effect on the void fraction distribution and the
wall shear stress is discussed in this section by examining the ability of gathering
bubbles to near the wall region and the wall shear stress under the similar liquid and gas
91
flow rate Jl and Jg.
For the upward turbulent bubbly flow with the wall peak of the void fraction, the
ratio γ between the maximum void fraction near the wall αgmax and that in duct center αgc
is defined as the peaking factor of the bubble gathering.
gcg /max . (5.7)
The larger γ indicates stronger bubble gathering near the wall. Generally, γ is
determined by the flow rate, bubble size, and duct geometry and so on. To show the
geometry effect, the flow conditions with the similar flow rates and bubble size were
chosen for the different experiments as shown in Table 5.3. Fig. 5.7 shows the
comparison of γ for the different duct geometries including small and large circular
ducts and noncircular ducts. The following results are obtained: (1) for the upward
turbulent bubbly flow in the circular ducts, γ for the small pipes γsc is larger than that for
the large pipes γlc, indicating stronger bubbles gathering ability in the smaller pipes; (2)
γln in the large noncircular duct is larger than that the large pipes γlc and especially near
the corner, where γln gets close to that for the small circular pipes γsc, indicating stronger
bubble gathering ability as compared to that in the large circular pipe due to the corner.
The duct geometry effect on the wall shear stress for two-phase flows is
examined by the well-known Lockhart-Martinelli correlation [5.9]. It shows the duct
geometry effect on the wall shear stress for the upward turbulent bubbly flow could be
estimated based on the single-phase flow under the same flow rate and a multiplier as
follows:
lwLtw ,
2
, , (5.8)
where 2
L is the frictional multiplier for two-phase flow and ηw,l is the single-phase wall
shear stress resulting from the liquid phase flow rate of Jl/αl. Next the duct geometry
effect on 2
L and ηw,l will be considered respectively to show its effect on the wall shear
stress for two-phase flows.
For the upward turbulent bubbly flow in the smooth pipe, according to
92
Chisholm-Larid correlation [5.9],
22 1211 XXL (5.9)
and gwlwX ,, / , (5.10)
where ηw,g is the single-phase wall shear stress resulting from the gas phase flow rate of
Jg/αg.
For the single-phase flow in the pipe, the wall shear stress ηw,l is usually
predicted based on the following empirical correlation:
2
, / 2w l l l lf W , (5.11)
where Wl is the mean flow rate of the single-phase flow; fl is the friction factor, and
could be estimated by fl=16/Rel and fl=0.0791/Rel0.25
for laminar and turbulent flow
respectively; Rel is Reynolds number, Rel=ρlVlDH/μl.
For the upward turbulent bubbly flow, substituting Eq. (5.11) into Eqs. (5.9) and
(5.10) for the liquid and gas phases to calculate the multiplier, it shows the duct
geometry has no effect on 2
L . Therefore, the duct geometry effect on the wall shear
stress for the upward turbulent bubbly flow could be estimated based on the
single-phase flow under the same flow rate where the larger duct corresponds to the
smaller wall shear stress. It means that the large square duct owns not only the stronger
bubble gathering ability to the wall and corner as compared with other large circular
ducts but also less wall shear stress compared with other smaller ducts. According to the
analysis in Section 5.2, it showed the flow with smaller wall shear stress, and steeper
void fraction profile is easier to form M-shaped velocity profile. Therefore, the
observation of the significant M-shaped Wl profile in the large square duct could be well
understood.
93
Table 5.3 Experimental conditions to for different geometries in the literatures. In the
table, C: circular duct; T: triangle duct; S: square duct; B: base length; H: height.
Exps.
Authors
Duct
shape
DH
(mm) Jl (m/s)
DB
(mm)
z
(DH)
Serizawa et al.[5.2]
C 60 1.03 3.5~4 20
Liu et al.[5.5]
C 57.2 0.75
1.083 2~4 36
Wang et al.[5.4] C 57.15 0.71
0.94 × 35
Shawkat et al.[5.6]
C 200 0.68 3~6 42
Sun et al.[5.7] C 101.6 1.021 4~5 18
Lahey et al.[5.8] T 50(B)
100(H) 1.0 5 73
Sun[5.1]
S 136 0.75
1.0 3~5 16
94
Fig. 5.7 Comparison of γ for different duct geometries for Jl around (a) 0.75m/s and (b)
1.0m/s. For noncircular geometry, ―-1‖ and ―-2‖ indicate data obtained near the wall
center and corner, respectively.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.351
2
3
4
5
6
7
8
9
c
Serizawa et al
Wang et al
Liu et al
Sun et al
Lopez et al-1
Sun et al-1
Lopez et al-2
Sun et al-2
Small circular pipe
Noncircular pipe
Large circular pipe
(b)
0 0.05 0.1 0.15 0.2 0.25 0.3 0.351
2
3
4
5
6
7
c
Wang et al
Liu et al
Shawkat et al
Sun et al-1
Sun et al-2
Small circular pipe
Large circular pipe
Noncircular pipe
(a)
95
5.5 Peak location of Wl near the wall
This section will discuss the peak location xl of the liquid phase velocity Wl near
the wall. At the peak location xl, the viscous force ηv is zero. To get xl, the lateral
distribution of ηv could be calculated on the bisector line by integrating of –F from the
wall or the duct center to the specified location x, respectively.
2/
2/0
H
H
D
xD
x
wv xdFxdF . (5.12)
For the fully-developed flow, ηv,D/2 = 0 due to the flow symmetry in the duct
center. Substituting Eqs. (5.1) and (5.3) to Eq. (5.12), and after rearranging, the lateral
distribution of ηv could be obtained as
1
the bubble blocking effect
2
1 2 / / 1
/ 2 ,
v w H l lm lm lm
l lm H
x D g
P g x D
, (5.13)
where αlm1 is mean void fraction between location x and wall; αlm2 is mean void fraction
between location x and center. The second term on the right hand shows the blocking
effect of bubble layer near the wall on wall retarding effect from transferring into the
core region. With M-shape velocity profile, for the fully developed flow, zero local
viscous force ηv is reached at the location xl due to the zero velocity gradients. According
to Eq. (5.13), at the location xl, we have 2 /lm lP g indicating it lies between
the wall and that of sign changing point of F. However, as shown from Figs. 5.2 and 5.3,
the velocity peak was between the sign changing point of F and the duct center, which is
not consistent with this analysis.
5.6 Flow characteristics in two layers
With the M-shaped liquid phase velocity profile, two layers with the different
flow types (pressure driven and shear driven flows) could be formed. In this section, the
flow characteristics in these two layers are presented based on the normalization of
96
liquid phase velocity.
Fig. 5.8 Definition of two layers.
5.6.1 Definition of two layers
Fig. 5.8 shows the definition of two layers divided by the peak of Wl. In the
inner layer where the wall strongly affects the entire region, the following
dimensionless quantities are defined,
lmWWWin /* , and mxxX in /* . (5.14)
The maximum magnitude of the velocity Wlm and the distance of this point to the
wall xm were chosen as the characteristic velocity and length for the different flow
conditions, respectively. As for the single-phase pipe flow driven by the constant
pressure gradient, the quadratic relation W*in =1- (X
*in -1)
2 could be derived in between
W*in and X
*in for the fully-developed laminar flow regime.
In the outer layer, due to symmetry at the center of the wall where the velocity
gradient is zero, there might exist an inflection point in the liquid velocity profile
between xm and xc as shown in Fig. 5.8. In the present study this feature will be not
focused due to the lack of the experimental data. In the outer layer where the wall
effects were blocked by the bubble layer, the following dimensionless quantities are
97
defined,
lclmlc WWWWWout *, and mc xxxxX out m
*. (5.15)
Herein the flow is driven by the local viscous shear generated by the faster inner
layer. The velocity difference (Wlm-Wc) and the distance (xc-xm) between the velocity
peak location and the duct center were chosen as the characteristic velocity and the
length for the different flow conditions, respectively. For the single-phase pipe flow
completely driven by the constant local net viscous shear dηv/ds (ηv: local viscous force
and s: coordinate direction, x or y), the quadratic relation W*out =(X
*out -1)
2 could be
derived in between W*out and X
*out for the fully-developed laminar flow regime. If the
flow is completely driven by the velocity difference like the Couette flow, the linear
relation W*out =1-X
*out could be derived for the fully-developed flow regime.
5.6.2 Velocity characteristics
By using the above normalization quantities and the measured experimental data
of the bubbly flow in the square duct [5.1], the velocity characteristics in the inner and
outer layers are shown along both the bisector and the diagonal lines in Fig. 5.9.
Because of the limitation of the pipe length in the experiments, the experimental cases
with the better developed M-shaped Wl profile was chosen. Besides since in Sun’s
experiments only a few points were measured in the inner layer, the cases of the upward
turbulent bubbly flow in the transition region were chosen. Considering the features of
the single-phase flow, it is treated all in the inner layer. In order to have an intuitive
understanding, the distributions of void fraction along both the bisector and the diagonal
lines could be found in Fig. 2.10 in Chapter 2.
In the inner layer, the velocity profile deviates from the quadratic function due to
the turbulent effects as shown in Figs. 5.9(a) and 5.9(b). Along the bisector line, it is
noted that W* under Jl=1.25m/s and Jg=0.27m/s becomes lager than the other cases. It
might be attributed to the following two reasons. Due to the high liquid superficial
velocity of this case, (1) the location of the velocity peak is closer to the wall; (2) the
98
flow might be less developed based on the uniform inlet flow distribution. In addition,
along the diagonal line near the corner it is observed that the velocity profile returns to
the quadratic function with the increase of the void fraction as shown in Fig 5.9(b). Two
possible reasons could be considered for this phenomenon. One is that the bubble
clustering region in these cases could be considered as a boundary which confined the
liquid flow between the wall and itself. It results in the smaller characteristic length as
compared with the single-phase flow. Therefore, the inertial effects in this layer will be
suppressed and the flow tends to be relaminarized. The other is that with the increase of
the void fraction, the flow in the inner layer also gets closer to the Couette flow, which
is completely driven by the velocity difference.
In the outer layer, the velocity profile is observed between the quadratic function
and the linear function as shown in Figs. 5.9(c) and 5.9(d). If much more bubbles are
accumulated near the wall and/or the corner as compared to the core region (steeper
void fraction distribution from the wall to the duct center), the velocity profile gets
closer to the Couette flow as shown in Fig. 5.9(d). It could be imagined that the
asymptote of the Couette flow could be reached for completely stratified flow of the gas
and liquid with the gas flow near the wall and the liquid flow in the duct center. In the
real situation, with the increase of the void fraction, the coalescence of the bubbles
occurs which makes the bubble migrate to the duct center region. Under the large
enough void fraction, considering the mixture of the fluids, the similar tendency to the
Couette flow will be formed between the duct center and near the wall region. Therefore,
it further confirms that for the upward turbulent bubbly flow in this large square duct,
two layers with the different flow types were easy to be formed. According to these
characteristics, the two-layer model which models the turbulent bubbly flow seperatedly
in different regions according to the different flow types was suggested to improve the
modeling of the upward turbulent bubbly flow therein.
In addition, unlike the flows in the circular pipes with the axisymmetric flow
distribution, for the single-phase turbulent flow in square duct there exists the secondary
flow in cross section due to the non-axisymmetric lateral turbulent intensities [5.10].
99
For the turbulent bubbly flow in the square duct, the non-axisymmetric void distribution
may also induce the horizontal secondary flow due to non-axisymmetric buoyancy
effect. The existence of the secondary flow could have a strong effect on flow
characteristics such as distorting the local mean velocity distribution. As shown in Fig.
5.9(d), it was observed that W*
out in the middle region, which might be related to the
convective transport of the axial liquid momentum due to the existence of secondary
flow. Therefore, in order to improve the model of the turbulent bubbly flow in the
non-circular ducts, the secondary flow effect should be considered. However, currently
there are no experimental information about the secondary flow of turbulent bubbly
flow therein, so the experimental measurements about the secondary flow is suggested
to carry out in the future work.
100
Fig. 5.9 Velocity characteristics in the inner layer along (a) the bisector line and (b) diagonal line; and in the outer layer (c) along the
bisector line and (d) the diagonal line. The mean void fraction could be approximated as αgm=Jg/(Jg+Jl).
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1(a)
Dimensionless Distance, X*
in
Dim
ensi
on
less
Liq
uid
velo
cit
y, W
* in
Symbol Jl, J
g
0.50 0.135
0.75 0.180
1.25 0.270
0.50 0.000
0.75 0.000
1.00 0.000
1.25 0.000
1-(X*-1)
2
turbulenteffect
Wall flowLaminar regime
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1(b)
Dimensionless Distance, X*
in
Dim
ensi
on
less
Liq
uid
velo
cit
y, W
* in
Symbol Jl, J
g
0.50 0.135
0.75 0.180
1.25 0.270
0.50 0.000
0.75 0.000
1.00 0.000
1.20 0.000
1-(X*-1)
2
Wall flowLaminar regime
Increase of
turbulenteffect
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1(c)
Dimensionless Distance, X*
out
Dim
ensi
on
less
Liq
uid
velo
cit
y, W
* out
Symbol J
l, J
g
1-X*
out
2
1-X*
out
0.75 0.090
0.75 0.135
0.75 0.180
1.00 0.090
1.00 0.135
1.00 0.180
1-Xout
*2
1-Xout
*
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1(d)
Dimensionless Distance, X*
out
Dim
ensi
on
less
Liq
uid
velo
cit
y, W
* out
Symbol J
l, J
g
1-X*
out
2
1-X*
out
Steeper
distribution
0.75 0.090
0.75 0.135
0.75 0.180
1.00 0.090
1.00 0.135
1.00 0.180
1-Xout
*2
1-Xout
*
Couette type flow
Couette
type flow
101
5.7 Numerical validation of the liquid phase velocity
In this section, the numerical validation of the liquid phase velocity profile was
carried out based on the experimentally measured pressure drop and the void fraction
distribution [5.1].
5.7.1 Numerical models
In order to obtain the lateral distribution of the axial liquid phase velocity W[,
the force balance equation in the axial direction is solved according to the formulation
proposed by Ishii and Hibiki [5.11]. For the gas phase:
, ,0.
g xz g g yz g
g g g DZ
dpg F
dz x y
(5.16)
For the liquid phase:
, ,1 1
1 0,g xz l g yz l
g DZ
dpF
dz x y
(5.17)
where dp/dz is the pressure gradient for the upward turbulent bubbly flow; ηxz,l, and ηxz,g
are the shear stress induced by the molecular and eddy viscosity for liquid phase and gas
phase, respectively; FDZ is the drag force due to the phase interaction; ρl and ρg are the
liquid and gas density, respectively; g is gravity acceleration: -9.81m2/s. By summing
Eqs. (5.16) and Eq. (5.17) and neglecting the gas phase due to ρg <<ρl,
, ,1 11 0.
g xz l g yz l
g l
dpg
dz x y
. (5.18)
Considering the geometrical symmetry, the governing equation on the bisector lines is:
11 0
L
g sz
g l
ddpg
dz ds
, (5.19)
where ds=dx or ds=dy for the bisector line vertical to y or x axis, respectively.
The one-equation turbulence model is used to express the liquid phase turbulent
102
Reynolds shear stress as ηsz,l=ρlvt dWl/ds in terms of the eddy viscosity vt and the liquid
phase velocity gradient dWl/ds. The turbulence eddy viscosity vt is expressed in the form
of t tv kl . k, lt and k are the turbulent kinetic energy, the characteristic length scale
and the velocity of turbulent structures for the upward turbulent bubbly flow,
respectively. Substituting the Reynolds stress ηsz,l to Eq. (5.19), it is obtained that
1
1 0
lg l t
g l
dWd kl
dp dsg
dz ds
. (5.20)
Eq. (5.20) was solved to get the liquid phase velocity profile Wl given by the
experimental measured pressure gradient, the void fraction distribution and the turbulent
kinetic energy. The flow continuity is maintained by setting the mean liquid flow rate.
For the upward turbulent bubbly flow, the turbulence could be induced by both
the liquid shear and the bubbles relative motion. When the bubble-induced turbulence
played a dominated role, the characteristic length scale lt could depend on the bubble
diameter DB and also the bubble-bubble interval distance lB-B due to the interactions
between the bubble-induced eddies. Here, the mean vertical bubble-bubble interval
distance lB-B is chosen as the flow characteristic length scale lt during solving Eq. (5.20).
As shown in Fig. 5.10, lB-B could be estimated based on the mean bubble diameter DB
and the void fraction αg as follows:
ggBN
BLBBB DNTWlB
111
. (5.21)
Fig. 5.10 Estimation of lB-B from the experiments. T, TB and TL are the total measuring
time, bubbles passing time and liquid passing time, respetively. WB is the mean bubble
velocity.
103
5.7.2 Numerical results
Fig. 5.11 shows the comparison between the predicted axial liquid phase
velocity Wl and the experimental measurement on the bisector line. It is observed that:
(1) with the characteristic length scale of lB-B, the liquid phase velocity profile Wl could
be predicted very close to the experimental measurement especially in core region,
indicating lB-B could be another suitable parameter to model the bubble-induced
turbulence for the upward turbulent bubbly flow. However, it still needs further
validations based on more experimental database. (2) As compared to the experimental
data, the predicted velocity profile Wl shows the flow all in the outer layer where the
flow was driven by shear. The lateral distribution of the mean viscous force ηv calculated
from the experiments confirms this point as shown in Fig. 5.12, since it is all negative.
The two assumptions are attributed for the disagreement with the experimental
measurements, that is, neglecting the flow acceleration and the uniform pressure
gradient distribution among the cross sections. However, in the experiments though the
void fraction distribution firstly reached the developed status at z = 4.5DH, the
liquid-phase velocity distribution was not confirmed whether it reached the
fully-developed status or still on the developing way at z = 16DH [5.1] since no lateral
downstream measurements were carried out. If the velocity was still on the developing
way, local flow acceleration term should be considered under the uniform pressure
gradient distribution. Therefore, to improve the upward turbulent bubbly flow modeling
in the noncircular ducts, it is necessary to establish the experimental database with
better developed-flow status to validate current models. Moreover the experiments with
the detailed flow developing process are also necessary to model the flow acceleration.
104
Fig. 5.11 Comparison between numerical predicted liquid phase velocity Wl and
experimental measurement on the bisector line for (a) Jl =0.75m/s and (b) Jl =1.0m/s,
respectively.
Fig. 5.12 Lateral distribution of mean viscous force ηv based on the experimental
measurements for (a) Jl =0.75m/s and (b) Jl =1.0m/s, respectively.
0 10 20 30 40 50 60 700.75
0.8
0.85
0.9
0.95
1
1.05
Distance from the wall x(mm)
Wl
Prediction (Jg=0.09)
Experimental data
Prediction (Jg=0.135)
Experimental data
Jl= 0.75m/s
(a)
0 10 20 30 40 50 60 701.05
1.1
1.15
1.2
1.25
1.3
1.35
Distance from the wall x(mm)W
l
Prediction (Jg=0.09)
Experimental data
Prediction (Jg=0.135)
Experimental data
(b)
Jl= 1.0m/s
0 10 20 30 40 50 60 70-15
-10
-5
0
Distance from the wall x (mm)
Vis
cou
s sh
ear
stre
ss
v
(a)
Jl = 0.75m/s
Jg (m/s)
0.09
0.135
0 10 20 30 40 50 60 70-20
-15
-10
-5
0
Distance from the wall x (mm)
Vis
cou
s sh
ear
stre
ss
v
(b)
Jl = 1.0m/s
Jg (m/s)
Jg0.09
Jg0.135
105
5.8 Summary and discussion
In this section, the formation of the typical M-shaped velocity profile of the
upward turbulent bubbly flow in the large square duct was studied by considering the
index of flow type, the formation conditions and the peak locations of the velocity
profile. The following findings were obtained therein.
(1) Neglecting the fluid acceleration and secondary flow, the local net driving
force F could be chosen as an index to show the flow type. Based on the sign
of F the formation of the M-shaped or W-shaped velocity profile could be
implied for the turbulent bubbly flow with the small and large void fractions
respectively, which has been verified for the small void fraction cases.
(2) According to the sign of F, it was found that for the turbulent bubbly flow in
the large square duct two layers with the different flow types could be
formed between the wall and the duct center. For the small void fraction
cases, two layers are the upward pressure driven flow near the wall, and the
viscous shear driven flow in the core region. For the large void fraction cases,
two layers are the upward pressure driven flow in the core region, and the
viscous shear driven flow near the wall.
(3) According to the analysis on the conditions for the formation of the
M-shaped velocity profile, the final flow type was determined by the lateral
void fraction distribution and the wall drag resistance, where the duct
geometry plays a dominated role on determining these two factors.
(4) The phenomenon of the M-shaped velocity profile in the large square duct
could be attributed to the geometry effect (a large square duct) which owns
not only stronger bubble gathering ability to the wall but also the smaller
wall shear stress.
(5) As for the peak location of the velocity, based on the current knowledge it
lies between the wall and that of sign changing point of F. However, the
experimental measurements of the velocity profile showed an inconsistent
106
result.
(6) According to the flow characteristics, the two-layer model for the upward
turbulent bubbly flow in the large square duct was suggested. The current
numerical validation indicates that the bubble-bubble distance could be
another important parameter to model the bubble-induced turbulence.
It is notable that the above analyses were carried out assuming the flow reached
the developed status without the secondary flow and uniform pressure gradient in the
cross section. Even though the experimental data at z = 16DH was used with better
developed status according to the analysis in Section 4.2, there are no further
measurements after z = 16DH to verify the flow reached the developed status or not. As
is known, the secondary flow among the cross section is the typical characteristics for
the single-phase turbulent in the noncircular ducts as compared to that in the circular
pipes. Here this effect was assumed to be neglected because of the following reason.
The experimental data of the lateral turbulent normal stresses (<uu> and <vv>) in [5.2]
showed that more symmetric as compared with that of the single-phase flow due to the
formation of the bubble layer near the wall and corner for the turbulent bubbly flow in
the square duct. It indicated the horizontal secondary flow may be very weak and
mainly exist in the region very near the corner; and the secondary flow could be
neglected. However, since there is no direct experimental information about the
horizontal convection and there may exist some error of the lateral normal turbulent
stresses (<uu> and <vv>), it still needs further consideration about this assumption.
Besides, the analysis implied the W-shaped velocity profile for the large void
fraction cases, the lack of the experimental data limited us the further validation. In
addition, the inconsistence about the peak location might be attributed to the
assumptions used. Therefore, in order to understand and develop the model of the
turbulent bubbly flow in the large square duct, the further detailed measurements should
be carried out with checking the flow developing status. The measurements of the
secondary flow were also suggested for the modeling of the turbulent bubbly flow in the
noncircular duct.
107
Reference
[5.1] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
[5.2] Serizawa, A., Kataoka, I., Michiyoshi, I., 1975, Turbulence structure of air-water
bubbly flow—II. local properties, Int. J. Multiphase Flow 2, 235–246.
[5.3] Theofanous, T. G., Sullivan, J., 1982, Turbulence in two-phase dispersed flows,
J. Fluid Mechanics 116, 343-362.
[5.4] Wang, S., Lee, S., Jones, O., Lahey, R., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
[5.5] Liu, T. J., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
pipe—I. liquid mean velocity and turbulence measurements, Int. J. Heat Mass
Transfer 36, 1049–1060.
[5.6] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2008, Bubble and liquid turbulence
characteristics of bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 34, 767-785.
[5.7] Sun, X., Smith, T. R., Kim, S., Ishii, M., Uhle, J., 2003, Interfacial structure of
air-water flow in a relatively large pipe. Exp. Fluids 34, 206-219.
[5.8] Lahey, Jr R. T., Bertodano, de M. L., Jones, Jr O. C., 1993, Phase distribution in
complex geometry conduits, Nucl. Eng, Des. 141, 177-201.
[5.9] Lockhart, R. W., Martinelli, R. C., 1949, ―Proposed Correlation of Data for
Isothermal Two Phase Flow, Two Component Flow in Pipes,‖ Chem. Eng. Prog.,
45, pp. 39–48.
[5.10] Hoagland, L. L., 1960, Fully Developed Turbulent Flow in Straight Rectangular
Ducts - Secondary Flow, Its Cause and Effect on the Primary Flow, Ph.D. thesis,
Department of Mechanical Engineering, MIT.
[5.11] Ishii, M., Hibiki, T., 2012, Thermo-Fluid Dynamics of Two-Phase Flow,
Springer.
108
109
Chapter 6
Void fraction of the turbulent bubbly flow
6.1 Introduction
As mentioned previously, for the turbulent bubbly flow, the deformation and
size of the bubble could significantly affect the bubbles’ behavior as well as the
associated momentum, heat and mass transfer processes. In the numerical predictions of
the turbulent bubbly flow, they are essential to calculate the bubble drag and lift forces,
which are two important forces to determine the fluid velocities, the phase distributions
and so on [6.1-6.4]. So far, a number of studies have been carried out on the effects of
the bubble’s deformation and size on the drag and lift force coefficients, CD and CL,
most of which were based on the single bubble experiments in different flow systems,
especially in the stagnant and simple shear flow systems [6.5-6.10]. For the bubbly flow,
the bubble size has long been an important measurement to understand the bubbly flow
characteristics with the aid of the optical or electrical probe techniques [6.11-6.13].
Using the general optical and electrical probe techniques, the time-averaged chord
length Lc of the bubbles could be measured, and indicate the characteristic axial size of
the bubbles. Based on the spherical bubble assumption, the volume equivalent spherical
bubble diameter (Sauter mean diameter: DSM ) was estimated by DSM =1.5Lc.
As for the measurement of the bubble’s deformation, the flow visualization is an
efficient way, and some correlations have been established based on the experiments of
the single bubble or droplet in the infinite stagnant liquids or various conduits [6.5].
However, since the flow visualization fails for the bubbly flows due to the low visibility,
there are few researches on the bubbles’ deformation for the bubbly flow except under
the low void fraction [6.14]. Currently, the numerical models of the bubbly flow mainly
110
consider the bubble deformation based on the single-bubble correlations [6.2, 6.8].
On the other hand, based on the double-sensor or multi-sensor probe, the
interfacial area concentration (IAC) ai , the void fraction αg, the averaged chord length
of the bubbles Lc could be measured for the bubbly flow simultaneously. Coupling the
IAC with the local time-averaged void fraction αg, the Sauter mean diameter DSM could
be given as follows:
DSM =6αg/ ai. (6.1)
With these three variables, it is regarded here that the bubble deformation for the
bubbly flow could be statistically reflected by combing the Sauter mean diameter DSM
and chord length Lc. In this chapter, the bubble deformation will be firstly studied in the
turbulent bubbly flow including many bubbles based on the recently established
experimental database by Sun [6.15]. After that, the numerical validation of the void
fraction will be carried out based on the experimental measurements.
6.2 Bubble deformation
6.2.1 Definition of bubbles’ statistical shape
To begin with, the time-averaged Sauter mean diameter DSM and chord length Lc
measured from the experiments were firstly shown in Figs. 6.1-6.3 as a database along
the bisector, diagonal and the near wall lines, respectively. The current analysis was
limited for the bubbles with the time-averaged DSM ranging from 3.5mm~5.5mm and Lc
ranging from 2.5mm~4.0mm. The following trends could be observed:
(1) When it gets close to the wall, Lc shows monotonously increase, but DSM
shows different tendency, indicating the effect of the bubble deformation changing from
the duct center to the near wall region. Since the present study mainly considers the
void fraction effects on the bubble deformation, the following comparisons will be
mainly carried out in the core region and neglecting the wall effects.
(2) DSM doesn’t change with increase of the gas flow rate or the void fraction,
111
but Lc clearly increases. According to the definition of Lc, the characteristic length of the
bubbles in the axial direction could be represented by Lc even for the different bubble
shapes [6.16]. Therefore, for the bubbly flow the change of Lc under the same DSM
implies the change of the bubble deformation statistically, which will be analyzed in
this section.
To investigate the bubble deformation for the bubbly flow, the shape factor of
the bubbles is defined and examined in this part. From the flow visualization for
different flow conditions, the bubbles could be assumed to be the ellipsoidal shape with
circular plane vertical to the axial direction from the statistical point of view, as shown
in Fig. 6.4. DV and DH are defined as the dimensions in the axial and horizontal
directions, respectively. Liu and Clark [6.16-6.17] have carried out a series of studies on
the relation between the chord length and the axial dimension of the bubbles for the
different shapes. Assuming a spherical or ellipsoidal bubble, DV could be calculated as
DV=1.5Lc if there were enough bubbles penetrated by the sensor.
So far, by a four-sensor optical probe, the void fraction, the IAC and the chord
length were measured. Assuming to the bubbles to be spherical, DSM could be calculated.
Similarly, assuming the bubbles to be statistically ellipsoids, it is possible to calculate
the bubble deformation based on DV=1.5Lc as follows.
Firstly, the shape of the bubbles is defined as the aspect ratio E between the
bubble horizontal and axial dimensions,
VH DDE / , (6.2)
Based on E, the volume VB and the surface area SB could be calculated by
66
322
VVHB
DEDDV
, (6.3)
and 23
732
22
12 EE
VB
EDS
. (6.4)
Then, the bubble deformation could be obtained by
112
23
7321
26
EEBV
B
SD
V
. (6.5)
According to the definition of DSM (iB
B
aS
V 66 ) and substituting DV = 1.5Lc, Eq.
(6.5) can be rearranged as
23
7321
2
5.1EEc
SM
L
Dr
. (6.6)
With DSM and Lc, E could be shown by the ratio between DSM and LC measured by the
four-sensor optical probe. After solving Eq. (6.6), it could be obtained as
2
422
624
843402
r
rrrE
. (6.7)
The relation between E and DSM/LC was plotted in Fig. 6.5. If DSM/LC is larger
than 1.5, it indicates E>1 which corresponds to the statistical oblate bubbles. If DSM/LC
equals to 1.5, it indicates E=1 which corresponds to the statistical spherical bubbles.
Otherwise, it indicates the statistical prolate bubbles. Therefore, the statistical bubble
shape could be summarized based on DSM/LC as
prolateFor 1.5
sphericalFor 5.1
oblateFor 5.1
c
SM
L
D (6.8)
In the following section, DSM/Lc will be examined to show the bubble statistical
deformation.
113
Fig. 6.1 Distributions of the Sauter mean diameter DSM and the chord length Lc along the bisector line under different flow conditions :(a)
Jl =0.5m/s; (b) Jl =0.75m/s; (c) Jl =1.0m/s; (d) Jl =1.25m/s. The solid and dashed lines show DSM, and Lc respectively.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(a) Jl =0.5 m/s
Jg0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(b) Jl =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(c)J
l =1.0 m/s
Jg
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(d)J
l =1.25 m/s
Jg
0.135 0.18 0.225 0.27
114
Fig. 6.2 Distributions of the Sauter mean diameter DSM and the chord length Lc along the diagonal line under different flow
conditions :(a) Jl =0.5m/s; (b) Jl =0.75m/s; (c) Jl =1.0m/s; (d) Jl =1.25m/s. The solid and dashed lines show DSM, and Lc respectively.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(a) Jl =0.5 m/s
Jg
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(b) Jl =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(c) Jl =1.0 m/s
Jg
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(d) Jl =1.25 m/s
Jg
0.135 0.18 0.225 0.27
115
Fig. 6.3 Distributions of the Sauter mean diameter DSM and the chord length Lc along the near-wall line under different flow
conditions :(a) Jl =0.5m/s; (b) Jl =0.75m/s; (c) Jl =1.0m/s; (d) Jl =1.25m/s. The solid and dashed lines show DSM, and Lc respectively.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(a) J
l =0.5 m/s
Jg
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(b) Jl =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(c)J
l =1.0 m/s
Jg
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.072
2.5
3
3.5
4
4.5
5
5.5x 10
-3
x (mm)
DS
M&
LC
(d) Jl =1.25 m/s
Jg
0.135 0.18 0.225 0.27
116
Fig. 6.4 Schematic diagram of the ellipsoidal bubble shape.
Fig. 6.5 Aspect ratio E vs. the ratio DSM/Lc for the ellipsoidal bubbles.
0 0.5 1 1.5 2 2.5 30
2
4
6
8
10
DSM
/Lc
E
Oblate bubble
Prolate bubble
Spherical bubble
117
6.2.2 Distribution of bubbles’ shape factor
Together with the void fraction distributions, Figs. 6.6-6.8 show DSM/Lc for
different flow conditions along the bisector, the diagonal and the near-wall lines,
respectively. It is observed as follows:
(1) DSM/Lc decreases when it gets close to the wall, and for the low void
fraction cases, DSM/Lc is greater than 1.5 in the core region and sometimes
less than 1.5 near the wall where there are high local void fraction and
liquid shear. According to the definition of DSM/Lc in Eq. (6.8), it means the
bubbles statistically behave oblate in the core region, which is qualitatively
in agreement with that observed in the single bubble or droplet experiments
in the infinite liquids [6.5].
(2) However, increasing the void fraction, DSM/Lc decreases in the whole area
including both the core region and near the wall, indicating that the bubbles
statistically behave more elongated. Especially for the largest void fraction
case here, the bubbles show statistically prolate even in the core region, and
the most prolate bubbles could be implied near the corner, which coincides
with the prolate bubbles cluster observed by the flow visualization.
The above analysis indicates that for the bubbly flow with the different flow
conditions the bubbles don’t always behave oblate even in the core region. In addition,
the existence of the many bubbles may affect the bubble deformation to deviate from
the single bubble case. However, before considering the void fraction effects, it is
necessary to examine whether the above observation is owing to the variation of the
bubble size for the different flow conditions or not. In the following, whether and how
the many bubbles affect the bubble deformation will be focused on.
118
Fig. 6.6 Distributions of the ratio r along the bisector line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75 m/s;
(c) Jl =1.0 m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(a) Jl =0.5 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(b) Jl =0.75 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(c) Jl =1.0 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(d) Jl =1.25 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.135 0.18 0.225 0.27
119
Fig. 6.7 Distributions of the ratio r along the diagonal line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75m/s;
(c) Jl =1.0m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(a) Jl =0.5 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(b) Jl =0.75 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(c) Jl =1.0 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(d) Jl =1.25 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.135 0.18 0.225 0.27
120
Fig. 6.8 Distributions of the ratio r along the near-wall line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75m/s;
(c) Jl =1.0m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(a) Jl =0.5 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(b) Jl =0.75 m/s
Jg
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(c) Jl =1.0 m/s
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
1
1.5
2
DS
M/L
C
(d) Jl =1.25 m/s
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070
0.1
0.2
x (mm)
Jg
0.135 0.18 0.225 0.27
121
6.2.3 Void fraction effects on shape factor
6.2.3.1 Bubble deformation for single-bubble case
As for the bubble deformation, theoretically it is governed by the force balance
on the bubbles from the continuous phase including the normal force, viscous shear
stress, gravity and the surface tension. The previous researches of the bubble
deformation in the single bubble or droplet experiments mainly considered the bubble
deformation correlated based on the dimensionless numbers including Eo, M, Re and
We, when it is dominated by different effects such as the surface tension, inertial or
viscosity effects.
For the bubble deformation of the ellipsoidal bubbles, Clift et al. [6.5]
recommended the correlation proposed by Wellek et al. [6.18] which analyzed
photographs of rising drops of immiscible liquids in continuous liquid. Therein, for low
M system with Eo < 40, the effect of Re is not important and the bubble deformation E0
for the single bubble could be estimated by
0.757
0 1 0.163H
V
DE Eo
D , (6.9)
or1/3
0H SMD D E . (6.10)
This correlation has been well validated based on the single-bubble experiments
[6.5, 6.19]. Currently, it is usually adopted in the two-fluid or multi-fluid model to
calculate the lift force coefficient CL which is related with the bubbles deformation and
is also crucial to the phase prediction of the numerical models [6.19]. Here, the bubble
size effect on the bubble deformation will be examined based on the above single
bubble correlation. By substituting Eq. (6.9) to Eq. (6.2) to eliminate the bubble size
effect, it is obtained that
122
0
1, other effects make the bubble more prolate;
/ 1, the same as the single bubble case;
1, other effects make the bubble more oblate.
R E E
(6.11)
Here, Eα is the statistical aspect ratio of the bubbles for the bubbly flow case.
Figs. 6.9-6.11 show the distribution of Eα/E0 along the bisector, the diagonal and the
near-wall lines, respectively. Based on the previous assumption and after eliminating
bubble size effect, it is observed that for the bubbly flow, the bubble deformation Eα was
over-predicted based on the single bubble correlation. Besides, with the increase of void
fraction, it deviates more from the single bubble correlation. Therefore, for the bubbly
flow, the correlation of the bubble deformation may need further correction considering
the void fraction effects.
123
Fig. 6.9 Distributions of the ratio Eα/E0 along the bisector line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75m/s;
(c) Jl =1.0m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(a)J
l =0.5 m/s
Jg
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(b) J
l =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(c)J
l =1.0 m/s
Jg0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(d)J
l =1.25 m/s
Jg
0.135 0.18 0.225 0.27
124
Fig. 6.10 Distributions of the ratio Eα/E0 along the diagonal line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75m/s;
(c) Jl =1.0m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(a)
Jl =0.5 m/s
Jg
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(b) Jl =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(c)
Jl =1.0 m/s
Jg
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(d) Jl =1.25 m/s
Jg
0.135 0.18 0.225 0.27
125
Fig. 6.11 Distributions of the ratio Eα/E0 along the near-wall line under different flow conditions: (a) Jl =0.5m/s; (b) Jl =0.75m/s;
(c) Jl =1.0m/s; (d) Jl =1.25m/s.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(a) Jl =0.5 m/s
Jg
0.045 0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(b) J
l =0.75 m/s
Jg
0.068 0.09 0.135
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(c) Jl =1.0 m/s
Jg
0.09 0.135 0.18 0.225
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.5
0.6
0.7
0.8
0.9
1
x (mm)
E/E
0
(d) Jl =1.25 m/s
Jg
0.135 0.18 0.225 0.27
126
6.2.3.2 Correction on the multi-bubble effect
As compared to the single bubble case with the similar bubble size, the
introduction of the many bubbles may change the force balance exerted on the bubbles
which dominates the bubble deformation. One effect of the many bubbles is to
statistically decrease the relative velocity VR due to the increase of the mixture viscosity
μm or drag on the bubbles [6.20-6.21]. Assuming the simple similarity between the
single bubble system and bubbly flow system, the decrease of VR and increase of μm
could modify the bubble deformation by decreasing the inertial effects and increasing
viscous effects respectively. The above two effects could be reflected by the Reynolds
number. In the present study, the void fraction effect on the statistical bubble
deformation was considered according to Re in the following form,
0 0
Re
Re
R
S RS
VE E f E f
V
, (6.12)
where Reα = ρlVRαDSM/μl and ReS=ρlVRSDSM/μl are Reynolds numbers for the bubbly
flow case with the void fraction α and for the single bubble case, respectively; and VRα
and VRS are the relative velocities between the bubbles and liquid phase for the bubbly
flow case and single bubble case, respectively.
Based on the assumption of the similarity between the single bubble system and
bubbly flow system, Ishii and Zuber [6.20] proposed the following relative velocity VRα,
75.11 RSR VV , (6.13)
According to the measurements of the buoyancy-driven bubbly flow, Garnier et
al. [6.21] proposed the following relation between VRα and α,
3/11 RSR VV . (6.14)
In the present study, Eq. (6.13) is chosen. Substituting Eq. (6.13) to Eq. (6.12)
and after rearranging, it is obtained that
-1fEE S . (6.15)
127
To identify the function f(1-α), Fig. 6.12 shows the relation between Eα/ES and
(1-α) by collecting the experimental data under the different flow conditions. A clear
tendency of the void fraction effect on the bubble deformation was shown and the
function f(1-α) is obtained as follows,
3.2-1-1 f . (6.16)
For the single bubble case α=0, the above Eq. (6.15) becomes Eq. (6.9) which keeps the
consistency.
Replacing ES in Eq. (6.9) by Eα in Eq. (6.15), therefore for the bubbly flow, the
statistical horizontal dimension of the bubbles in Eq. (6.9) could be calculated by,
1/3 0.7670.7571 0.163Eo 1-H SMD D . (6.17)
6.2.4 Bubble size validation
Figs. 6.13 and 6.14 show the comparisons among the statistical horizontal
dimensions of the bubbles calculated from the experimental data DH.exp, DHS calculated
based on Eq. (6.9) and DHα calculated based on Eq. (6.17). It indicates that with ES, the
bubble deformations for the bubbly flow are all over-predicted, but with Eα the bubble
deformation for the bubbly flow could be predicted much better. Therefore, Eq. (6.17) is
recommended to calculate the statistical horizontal dimension DH for the bubbly flow.
However, it is notable that the above correlation Eqs. (6.15)-(6.17) was validated only
for the limited range of the bubble size around 3.5mm~5.5mm and the void fraction
around 5%~30%. For the bubbly flow with the different range of the bubble size and
void fraction, it still needs further validation and consideration based on more
experimental data in the future.
128
Fig. 6.12 Ratio R=Eα/E0 vs. the liquid fraction (1-α) under all the tested flow conditions.
Fig. 6.13 DH.exp vs. DHS under all the tested flow conditions.
0.65 0.7 0.75 0.8 0.85 0.9 0.95 10.4
0.5
0.6
0.7
0.8
0.9
1
liquid fraction (1-)
R
experimental data
(1-)1.6
(1-)2.3
(1-)3
3 3.5 4 4.5 5 5.5 6
x 10-3
3
3.5
4
4.5
5
5.5
6x 10
-3
DHS
DH
,ex
p
experimental data
DH,exp
=DHS
DH,exp
=0.8DHS
129
Fig. 6.14 DH.exp vs. DHα under all the tested flow conditions.
6.3 Numerical validation of the void fraction
In this section, the numerical validation of the lateral distribution of the void
fraction will be carried out based on two-fluid model and the experimentally measured
turbulent intensities and the liquid-phase velocity distributions [6.15].
6.3.1 Numerical models
As shown in Chapter 5, in order to validate the liquid-phase velocity distribution
the force balance in the axial direction was solved based on the mixing length theory to
model the turbulent Reynolds shear stress. In this chapter, the force balance equation in
the horizontal direction will be solved in order to obtain the lateral distribution of the
void fraction following the method used in the small circular pipe [6.22]. Firstly,
assuming the fully developed flow status and neglecting the secondary flow effect, the
force balance in the horizontal direction is shown for the liquid and gas phases,
respectively. For the gas phase:
3 3.5 4 4.5 5 5.5 6
x 10-3
3
3.5
4
4.5
5
5.5
6x 10
-3
DH
DH
,ex
p
experimental data
DH,exp
=1.05DH
DH,exp
=DH
DH,exp
=0.95DH
130
0
G G
xx xy
Lx
P
X X Y
τ τM , (6.18)
0
G G
xy yy
Ly
P
Y X Y
τ τM . (6.19)
For the liquid phase:
1 1
1 0
L L
xx xy
Lx
P
X X Y
τ τM , (6.20)
1 1
1 0
L L
xy yy
Ly
P
Y X Y
τ τM , (6.21)
where MLx and MLy are the lift forces in X and Y directions, respectively. Eliminating
the pressure gradient by 20.6118.6 and 21.6119.6
and neglecting the turbulent stress of the gas phase due to 1LG , we have
1 10
L L
xx xy Lx
X Y
τ τ M, (6.22)
1 1
0
L L
xy yy Ly
X Y
τ τ M. (6.23)
With the above assumptions, the local void fraction distribution is determined by
the balance between the lateral turbulence dispersion force and the lift force. The lateral
turbulence dispersion is closely related to turbulence anisotropy which will be
considered in Chapter 7. Currently, the experimental measurements of the lateral
turbulence will be adopted.
Considering the geometrical symmetry, on the bisector and diagonal lines, Eqs.
(6.22) and (6.23) could be simplified as follows:
1
0
L
xx Lxd
dX
τ M =>
XGXF
dX
d
1
1 (6.24)
131
where duu
F XuudX
and L R LC W dWG X
uudX .
By solving Eq. (6.24) the lateral void fraction distribution could be obtained as
''
Gc0 0 0
Gc
1- 1- '' exp ' ' '' exp ' '
1- 1-
X y X
L R LcL LC
Lc
X G y F y dy dy F y dy
C W uuW W
uu uu
(6.25)
where αgc, LCW and Lcuu are the void fraction, liquid phase velocity, the normal
turbulent intensities at the duct center.
To solve Eq. (6.25), the lateral distributions of the liquid phase velocity and the
normal turbulent intensities are required in addition to αgc according to the experimental
databases. It is notable that in [6.15], the experimental measurements were carried out
based on the coordinate parallel with the wall direction. Here, during solving Eq. (6.25)
along the diagonal line, the relation about the turbulent normal components shown Fig.
6.15 is adopted.
The relative velocity WR is determined by the following correlation
4
3
l g
R
l
dW
g
. (6.26)
As for the lift coefficient CL, Tomiyama’s correlation [6.8] was used and therein
the bubble horizontal dimension DH was considered by 3/1757.0
0 163.01 EoDD SMH
and 767.03/1757.0 1163.01 EoDD SMH , respectively.
6.3.2 Numerical results
Firstly, Figs. 6.16-6.18 show the comparison of the lateral void fraction
distribution predicted by Eq. (6.25) and measured experiments at z=16DH along the
bisector and diagonal lines respectively. From these figures, it could be observed that:
(1) The predictions of the void fractions are in well consistent with the
132
experimental measurements for the cases (Jl =0.75m/s, Jg =0.09m/s, Jl =0.75m/s, Jg
=0.135m/s and Jl =1.0m/s, Jg =0.135m/s). However, for the other cases, the prediction
showed quite different behaviors.
(2) Based on the current experimental data, the choice of the bubble horizontal
dimension slightly modified the prediction of the void fraction.
The following characteristics are considered to be responsible of the above
observations. (1) According the developing status as analyzed in Fig. 4.9 of Chapter 4,
for the cases (Jl =0.75m/s, Jg =0.09m/s, Jl =0.75m/s, Jg =0.135m/s and Jl =1.0m/s, Jg
=0.135m/s), the flow showed better developed status at the location z=16DH. However,
for the other cases the liquid phase velocity was not either lifted up for the low void
fraction cases or entered the third developing stage for the large void fraction cases,
which deviates from the assumptions of Eq. (6.25). (2) The current bubble size is
around 3.5mm~5.5mm, among which there is little effect of DH0 or DHα on the lift
coefficients CL.
Fig. 6.15 Schematic diagram of the turbulent components along the diagonal line.
133
Fig. 6.16 Comparison between numerical predicted void fraction α and experimental measurement for Jl =0.5m/s at z=16DH:
with Jg =0.09m/s (a) along the bisector line; (b) along the diagonal line;
with Jg =0.135m/s (c) along the bisector line; (d) along the diagonal line.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.5 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.5 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.5 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)v
oid
fracti
on
Jl=0.5 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
(a) (b)
(c) (d)
134
Fig. 6.17 Comparison between numerical predicted void fraction α and experimental measurement for Jl =0.75m/s at z=16DH:
with Jg =0.09m/s (a) along the bisector line; (b) along the diagonal line; with Jg =0.135m/s (c) along the bisector line; (d) along the
diagonal line; with Jg =0.18m/s (e) along the bisector line; (f) along the diagonal line.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.18
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=0.75 J
g=0.18
experimental data
Predicted with DH
Predicted with DH0
(a)
(b)
(e) (c)
(d) (f)
135
Fig. 6.18 Comparison between numerical predicted void fraction α and experimental measurement for Jl =1.00m/s at z=16DH:
with Jg =0.09m/s (a) along the bisector line; (b) along the diagonal line; with Jg =0.135m/s (c) along the bisector line; (d) along the
diagonal line; with Jg =0.18m/s (e) along the bisector line; (f) along the diagonal line.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.18
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.09
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.135
experimental data
Predicted with DH
Predicted with DH0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.070.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
Distance from the wall x, (m)
void
fracti
on
Jl=1.0 J
g=0.18
experimental data
Predicted with DH
Predicted with DH0
(a)
(b)
(e) (c)
(d) (f)
136
6.4 Summary and Discussion
In this chapter, the void fraction effect on the bubble deformation was firstly
studied. Numerical validation of the void fraction was then carried out based on the
experimental results. The following findings were obtained:
(1) For the bubbly flow, the existence of the many bubbles could affect the
statistical bubbles’ deformation. Compared to the single-bubble case, the
bubbles behaved more prolate with the increase of the void fraction.
(2) Considering the void fraction effect, a new correlation for the statistical
bubble deformation were proposed and validated combining the bubble size
and void fraction.
(3) The numerical validation showed that only for the cases with the
better-developed flow, the predictions of the void fractions were in well
consistent with the experimental measurements.
It is notable that the analysis on the void fraction effect on the bubble
deformation was based on some assumptions such as all the ellipsoidal bubble shape
and enough bubble number during the measurements. Since the measurements were
carried out in the large duct, the wall effect was also neglected. Whether these
assumptions are suitable or not still needs further validation based on more
experimental data in the future. In addition, currently the new correlation of the bubble
deformation was proposed and validated for the mean bubble size around
3.5mm~5.5mm and the limited void fraction range. For the wider range of the bubble
size and flow conditions, it still needs further study in the future. The numerical work
further suggests the importance of knowing the developing status for the modeling of
the turbulent bubbly flow.
137
Reference
[6.1] Ishii, M., Hibiki, T., Thermo-Fluid Dynamics of Two-phase flow, Second
edition, 2010, Springer.
[6.2] Tomiyama, A., Shimada, N., 2001, A numerical method for bubbly flow
simulation based on a many-fluid model, Trans. ASME, J. of Pressure Vessel
Tech. 23, 510-516.
[6.3] Bunner, B., Tryggvason, G., 2003, Effect of bubble deformation on the
properties of bubbly flows, J. Fluid Mech. 495, 77–118.
[6.4] Lu, J. C., Tryggvason, G., 2008, Effect of bubble deformability in turbulent
bubbly upflow in a vertical channel, Phys. Fluids 20, 040701.
[6.5] Clift, R., Grace, J.R., Weber M.E., 1972, Bubbles, drops, and particles, Science.
[6.6] Tomiyama, A., Kataoka, I., Zun, I., and Sakaguchi T, 1998, Drag coefficients of
single air bubbles under normal and micro gravity conditions, JSME Int. J., Ser.
B 41, 472
[6.7] Magnaudet, J., Eames, I., 2000, The motion of high-Reynolds number bubbles
in homogeneous flows, Annual Rev. Fluid Mech. 32, 659-708.
[6.8] Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002, Transverse migration of
single bubbles in simple shear flows, Chem. Eng. Sci. 57, 1849-1858.
[6.9] Legendre, D., 2007, On the relation between the drag and the vorticity produced
on a clean bubble, Phys. Fluids 19, 018102.
[6.10] Rastello, M., Mari´e, J. L., Lance, M., 2011, Drag and lift forces on clean
spherical and ellipsoidal bubbles in a solid body rotating flow, J. Fluid Mech.
682, 434.
[6.11] Serizawa, A., Kataoka, I., Michiyoshi, I., 1975, Turbulence structures of
air-water bubbly flow-I. measuring techniques, Int. J. Multiphase Flow 2,
221-233.
[6.12] Liu, T. J., 1993, Bubble size and entrance effects on void development in a
vertical channel, Int. J. Multiphase Flow 19, 99-113.
138
[6.13] Liu, T. G., Bankoff, S. G., 1993, Structure of air-water bubbly flow in a vertical
pipe—II. Void fraction, bubble velocity and bubble size distribution, Int. J. Heat
and Mass Transfer 36, 1061–1072.
[6.14] Fujiwara, A., Danmoto, Y., Hishida, K., Maeda, M., 2004, Bubble deformation
and flow structure measured by double shadow images and PIV/LIF, Exp.
Fluids 36, 157–165.
[6.15] Sun, H. M., 2014, Study on upward air-water two-phase turbulent flow
characteristics in a vertical large square duct, Doctor Thesis, Kyoto University.
[6.16] Clark, N. N., Turton R, 1988, Chord length distributions related to bubble size
distributions in manyphase flows, Int. J. Multiphase Flow 14, 413-424.
[6.17] Liu, W., Clark, N. N., 1995, Relationships between distributions of chord length
and distributions of bubble sizes including their statistical parameters, Int. J.
Multiphase Flow 21, 1073-1089.
[6.18] Wellek, R. M., Agrawal, A. K., Skelland, A, H, P, 1966, Shape of liquid drops
moving in liquid media, AIChE J. 12, 854-862.
[6.19] Tomiyama, A., Celata, G. P., Hosokawa, S., Yoshida, S., 2002, Terminal
velocity of single bubbles in surface tension force dominant regime, Int. J.
Multiphase Flow 28, 1497-1519.
[6.20] Ishii, M., Zuber, N., 1979, Drag coefficient and relative velocity in bubbly,
droplet or particulate flows, AIChE J. 25, 843-855.
[6.21] Garnier, C., Lance, M., Mariém J. L., 2002, Measurement of local flow
characteristics in buoyancy-driven bubbly flow at high void fraction,
Experimental Therm. Fluid Sci. 26, 811-815.
[6.22] Wang, S., Lee, S., Jones, O., Lahey, Jr R. T., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
139
Chapter 7
Turbulence modulation of the turbulent bubbly flow
7.1 Introduction
As compared with the single-phase turbulent flow, for the turbulent bubbly flow
the presence of the bubbles will modulate the turbulence characteristics by the following
mechanisms: (1) Altering the shear-induced turbulence by modifying mean flow
properties such as the mean shear rate; (2) Inducing additional pseudo turbulence by
agitating the flow such as wake structure behind the bubbles; (3) Interacting with the
existing shear-induced eddies by being captured or colliding; (4) Inducing additional
interaction between bubble-induced and shear-induced eddies. Due to the above
mechanisms, the turbulence modulation such as turbulence enhancement or suppression
occurs in the turbulent bubbly flow, which has been partially discussed in [7.1].
Though several studies on the turbulence modulation [7.2-7.13] have been
carried out for the turbulent bubbly flow, the understanding and modeling are still
unsatisfactory so far. In this Chapter, the turbulence modulation of the upward turbulent
bubbly flow will be further studied based on the recently established experimental
database. It will aim at understanding the following questions:
(1) How much turbulent energy was induced by the bubbles?
(2) How does the interaction between the bubble-induced turbulence and the
wall-induced turbulence behave?
(3) What is the property of the bubble-induced turbulence such as the
turbulence anisotropy?
(4) How to model the bubble induced turbulence considering the above
properties?
140
7.2 Turbulent kinetic energy
For the turbulent bubbly flow in the ducts, the turbulence generation includes the
wall-induced turbulence tkw which is the main mechanism for the single-phase turbulent
flow and the bubble-induced turbulence tkb. Several studies [7.3-7.4, 7.8-7.11] have
been conducted to model the bubble-induced turbulence by comparing the turbulence
with and without the bubbles. The theoretical analysis could be found in [7.4, 7.11],
which estimated the bubble-induced turbulence based on the potential theory [7.4] and
based on the mean flow velocity and the wall shear stress [7.11], respectively. The
uniform bubbly flow experiments in [7.9] showed that below the critical void fraction αc
of the order of 1%, the bubble-induced pseudo turbulence could be well estimated by
the potential flow model, and above the critical void fraction αc, the turbulence was
strongly amplified by the hydrodynamic interactions between the bubbles and also the
background turbulence. The DNS study [7.14] showed a large difference from the
prediction by potential flow model.
This part focuses on the question of how much turbulent energy could be
induced by the bubbles. The turbulence databases shown in Chapter 2 will be adopted.
For the turbulent bubbly flow in the ducts, the wall-induced turbulence tkw and the
bubble-induced turbulence tkb are coupled with each other. Therefore, the knowledge of
both the solely wall-induced turbulence and the solely bubble-induced turbulence are
required to understand the final turbulent characteristics therein. For the solely
wall-induced turbulence, numerous researchers has been carried out for the single-phase
flow, based on which we reached a comprehensive understanding. However, for the
solely bubble-induced turbulence, there are still very few studies except the experiments
by Lance and Bataille about the uniform bubbly flow for the lower void fractions lower
than 3% [7.9]. For cases with larger void fraction, the solely bubble-induced turbulence
is different due to the stronger effect of the bubble-bubble interaction. To quantify the
solely bubble-induced turbulence for a wider range of void fraction, the recently
established experimental database of the bubbly flow in the ducts will be used and
141
analyzed in the following. Firstly the comparison between the bubble-induced
turbulence tkb and wall-induced turbulence tkw will be carried out to understand the role
of the bubble-induced turbulence. After that, the relations between the bubble-induced
turbulence tkb, the void fraction α and wall-induced turbulence tkw will be considered.
Finally, the interaction between bubble-induced turbulence and wall induced turbulence
will be discussed based on their relations.
The role of the bubble-induced turbulence here is considered by the ratio of the
turbulent kinetic energies between the turbulent bubbly flow tkt and the single-phase
flow tks under the same flow rate according to existed data, which were shown in Figs.
7.1-7.5 for the circular pipes with small and large diameter and also the noncircular
ducts. Larger ratio tkt / tks indicates stronger role of the bubble-induced turbulence. Here,
generally two-sets of the cases with closer liquid flow rates are given for simplicity and
comparison. Similar tendency could be found for the other cases. As shown from Figs.
7.1-7.5, the following trends could be obtained:
(i) Void fraction and liquid flow rate effect
The ratio tkt / tks increases with the increase of the void fraction for the same
liquid flow rate, and shows larger for cases with the smaller liquid flow rate for the
similar void fraction. It is easy to understand this observation because of the increase of
the bubble-induced turbulence with increasing the void fraction and the decrease of the
wall-induced turbulence with decreasing the liquid flow rate.
(ii) Turbulence enhancement or reduction
Generally, the ratio tkt / tks is greater than 1 especially in the duct center of the
large duct indicating the turbulence enhancement by the bubbles, except several cases in
the small circular pipe with very small gas flow rate and high liquid flow rate where the
turbulence reduction could be observed in the near wall region such as Jl=1.03m/s and
X%=0.0085 in Serizawa’s experimental database [7.2-7.3]. The turbulence reduction
under these cases was discussed in [7.8] and will be considered later in this chapter.
142
(iii) Lateral distribution
Except two cases under Jl=0.491m/s in Hibiki’s experimental database [7.15]
which were in the transition regime, generally the ratio tkt / tks for the turbulent bubbly
flow decrease from the duct center to the wall even with much higher void fraction near
the wall. It indicates the bubble-induced turbulence played a dominated role in the duct
center as compared with the near wall region. The exception of the two cases mentioned
above might be attributed to the bubble coalescence and breakup which induced another
mechanism for the turbulence generation and will not be considered here.
(iv) Geometry effect
Fig. 7.6(a) summarizes the ratio <ww>tc/<ww>sc of the axial turbulent intensity
for the turbulent bubbly flow in the duct center <ww>tc to that of the single-phase
turbulence <ww>sc for different flow conditions and ducts. It shows the larger ratio
tkt/tks could be observed for the larger duct size DH. Fig. 7.6(b) shows the contour lines
for the flow conditions Jl and Jg under which the ratio <ww>tc/<ww>sc is around 5 and 25,
respectively. Here the flow conditions on the contour lines were summarized and
interpolated from the corresponding references for different ducts. It was found that the
contour lines for the small circular pipes from Hibiki’s database [7.15] lie in the upper
region of that for the large ducts from Shakwat’s [7.13] and Sun’s database [7.16]. For
the Serizawa’s database [7.2] in the small circular pipe, the critical conditions for
<ww>tc/<ww>sc ≈ 25 were not reached even in the slug regime. It means that under the
same flow condition, the ratio <ww>tc/<ww>sc in the larger ducts is larger than that in the
small ducts. For example, for the case of Jl ≈1.0m/s and Jg≈0.1m/s as shown in
Fig.7.6(b), <ww>tc/<ww>sc ≈10 in the large ducts while <ww>tc/<ww>sc ≈5 in the small
ducts. Therefore, the bubble-induced turbulence played a more dominate role in the
larger ducts especially in the core region. The following two factors might be
responsible for this observation (a) more space for the bubble’s horizontal deformation
and wake development behind the bubbles for the large ducts which induced stronger
turbulence; (b) less effect from the wall-induced turbulence for the large ducts.
143
Fig. 7.1 Lateral distributions of the ratio of the turbulent kinetic energies between tkt and
tks in the small circular pipe with DH=60mm based on Serizawa’s experimental database
under (a) Jl =0.74m/s and (b) Jl =1.03m/s [7.2].
Fig. 7.2 Lateral distributions of the ratio of the turbulent kinetic energies between tkt and
tks in the small circular pipe with DH=50.8mm based on Hibiki’s experimental database:
under (a) Jl =0.491m/s and (b) Jl =0.986m/s [7.15].
0
5
10
15
r/R
t kt/t
ks
Jl=0.74m/s X %
0.000
0.0119
0.024
0.0358
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0
5
10
15
r/R
t kt/t
ks
Jl=1.03m/s
X %0.000
0.0085
0.0170
0.0258
0.0341
0.0427
0.0511
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
20
40
60
80
r/R
t kt/t
ks
J
l=0.491m/s J
g
0.000
0.0275
0.0556
0.129
0.190
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
10
20
30
40
r/R
t kt/t
ks
J
l=0.986m/s J
g
0.000
0.0473
0.113
0.242
0.321
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
(a) (b)
(a) (b)
144
Fig. 7.3 Lateral distributions of the ratio of the turbulent kinetic energies between tkt and
tks in the small circular pipe with DH=60mm based on Liu’s experimental database [7.7].
Fig. 7.4 Lateral distributions of the ratio of the turbulent kinetic energies between tkt and
tks in the large circular pipe with DH=200mm based on Shakwat’s experimental database
under (a) Jl =0.45m/s and (b) Jl =0.68m/s [7.13].
0 0.2 0.4 0.6 0.8 10
20
40
60
r/R
t kt/t
ks
J
l J
g
0.376 0.00
0.376 0.18
0.753 0.18
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
r/R
0
20
40
60
80
r/R
t kt/t
ks
Jl=0.45m/s J
g
0.0000.015
0.0850.100
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
r/R
0
20
40
60
r/R
t kt/t
ks
J
l=0.68m/s J
g
0.0000.015
0.0500.085
0.1000.180
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
r/R
(a) (b)
145
Fig. 7.5 Lateral distributions of the ratio of the turbulent kinetic energies between tkt and tks in the large circular pipe with DH=136mm
based on Sun’s experimental database: under Jl =0.75m/s (a) along the bisector line; (b) along the diagonal line;
under Jl =1.0m/s (c) along the bisector line; (b) along the diagonal line [7.16].
0
20
40
60
t kt/t
ks
Jl=0.75m/s Bisector
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 1
0.05
0.1
0.15
0.2
0.25
2x/DH
0
20
40
60
t kt/t
ks
J
l=0.75m/s Diagonal
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 1
0.05
0.1
0.15
0.2
0.25
2x/DH
0
10
20
30
t kt/t
ks
Jl=1.0m/s Bisector
0 0.2 0.4 0.6 0.8 1
0.05
0.1
0.15
0.2
0.25
2x/DH
0.0000.0900.135
0.1800.225
0
10
20
30
t kt/t
ks
J
l=1.0m/s Diagonal
0.0000.0900.135
0.1800.225
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
2x/DH
(a) (b)
(c) (d)
Jg Jg
Jg
Jg
146
Fig. 7.6 (a) The relation between the ratio <ww>tc/<ww>sc and the local void fraction α in
the duct center for different ducts; (b) The contour lines for the flow conditions Jl and Jg
when <ww>tc/<ww>sc ≈ 5, 10, 25.
0 0.05 0.1 0.15 0.2 0.25 0.30
50
100
150
c
<w
w>
tc /
<w
w>
sc
Liu
Serizawa
Sun
Shawkat
(a)
(b)
147
7.2.1 Bubble-induced turbulence at the duct center
Based on the above analysis, in the following the turbulent intensities in the duct
center will be mainly considered and analyzed especially in the large ducts where the
bubble-induced turbulence dominates the local turbulence in order to quantify and
model the bubble-induced turbulence. Here, we consider the turbulence in the duct
center as the solely bubble-induced turbulence since therein the shear-induced
turbulence is very weak due to the flat liquid phase velocity profile or small liquid phase
velocity shear. In addition, we also neglect the horizontal diffusion and convective
effect which are extremely weak in the core region. In other words, the turbulence in the
core region was mainly induced because of the bubble agitation.
Fig. 7.7 shows the relation between the axial turbulent intensities <ww>c and the
local void fraction α in the duct center based on the above experimental databases.
Besides, three correlations for the bubble-induced turbulence were also shown in Fig.
7.7, which are 22.0 rW proposed by [7.9, 7.11] based on the potential theory and
experimental results for α < 1%, 2
rW proposed by [7.14] by fitting their DNS results
for α<24%, and 6.172.0 proposed by [7.5] by fitting their experimental results
including the near wall region for α>1.3%. Wr is the relative velocity between the
bubbles and liquid phase and Wr=0.23m/s (the terminal velocity) in the above
correlations. As shown from Fig. 7.7(a), the following trends could be obtained:
(i) Duct geometry effect
Under the same void fraction, the bubble-induced turbulence is stronger in the
large ducts as compared to that in the small circular pipes. The discrepancy increases
with the increase of the void fraction. Similarly it could also be attributed to the
suppression of the bubble’s horizontal deformation and wake development in the small
circular pipes.
(ii) Liquid flow rate effect
148
The turbulence for the turbulent bubbly flow in the large ducts shows much
more scattered than that in the small circular pipes. The scattering in the large ducts was
found to be related with different liquid flow rates, which is consistent with [7.9].
Fig.7.7 (b) shows the ratio <ww>/Wl for the above cases. It was found that after dividing
by the local liquid velocity Wl or Reynolds number Rel, the scattering could be
significantly improved especially in the large duct. In addition, based on <ww>/Wl the
dependence of the turbulence on the duct size could be clearly shown especially for the
cases with α>5%. The deviation for further higher void fraction could be attributed to
that the flow was in the transition regime where bubble-coalescence is needed to be
considered. The ratio <ww>/Wl shows linear almost linear increase with the void
fraction α, where the slope should be determined by the duct size DH, the bubble size DB,
the bubble deformation E, and the relative velocity of the bubbles Wr. However, how it
relates to these parameters, it still needs further study in the future.
(iii) Correlation validation of 22.0 rW
Generally, the bubble-induced turbulence is larger than the predictions by
22.0 rW especially in the large ducts, which indicates the important role of the wake
behind the bubbles to the turbulence generation. However, in the small circular pipe, the
discrepancy between the potential theory and experimental measurements is apparently
smaller than that in the larger ducts. It is noticed that in the two-fluid models proposed
by Lahey [7.17], Bertodano et al. [7.18], Politano et al. [7.19], and Hosokawa and
Tomiyama [7.20], the bubble-induced turbulence was predicted by this potential theory
and validated based on the experimental database in the small ducts. The predictions
therein were basically in consistent with the experimental measurements. However, for
the large pipes the prediction shows larger discrepancy, and the discrepancy increases
with the void fraction α. Therefore, the new model considering the bubble-induced
wake and bubble deformation is required for the turbulent bubbly flow in the large ducts
for more accurate turbulence prediction therein.
149
(iv) Correlation validation of 2
rW
The turbulence because of the wake behind bubbles was included in the
correlation 2
rW [7.14] by fitting their DNS results with tens of the bubbles in a
limited domain. The prediction shows larger than the experimental measurements in the
small circular pipes but still smaller than that in the larger circular and noncircular ducts,
which might be attributed to the limited bubble number and computational domain.
(v) Correlation validation of 6.172.0
The correlation 6.172.0 was proposed in [7.5] by directly extracting the
wall-induced single-phase turbulence from the total turbulence of the turbulent bubbly
flow even the whole flow region. Though the prediction shows closer to the
experimental results in the large ducts, it is lack of the physical considerations and the
scattering phenomenon of the turbulence for different liquid flow rates could not be
predicted.
In addition, all the above correlations are independent of the bubble size and
deformation which played important roles in the bubble-induced turbulence. It made the
predictions deviate more from the experimental measurements in the large ducts.
In [7.21], based on the plausible assumption that all the energy lost because of
the drag resistance drag
LF on the bubbles was converted to the turbulent kinetic energy
in the bubble wake, the additional source term B
LS in the liquid phase turbulent kinetic
energy tk transport equation was proposed for the bubbles contribution as follows:
rL
B
LS VF drag. (7.1)
However, it is notable that for the turbulent bubbly flow the work done by the
drag force drag
LF contributes not only to the turbulent kinetic energy tk but also the
mean kinetic energy tM, where the total energy in the flow is kMtotal ttt . Hence, in
this way the double computation of the mean energy occurs in the governing equations
150
of the bubbly flow. This was mentioned and discussed by Yokomine et al. for the
turbulent flow with the dispersed solid particles [7.22]. Therein the double computation
was extracted from the source term by modifying the drag coefficient as follows:
StokesDD
W
D CCC , (7.2)
where StokesDC , is the drag coefficient for the dispersed particles based on the Stokes
flow.
In the following, this method proposed by Yokomine et al. [7.22] will be
adopted. For simplicity, the energy balance for the liquid phase turbulence kinetic
energy only at the duct center will only be considered where the shear contribution, and
horizontal diffusion and convection could be neglected. Therein, the turbulent kinetic
energy tkc was determined by the local bubble contribution B
LS and dissipation rate L
as follows.
0drag LrLL
B
LS VF . (7.3)
For the turbulent bubbly flow, the dissipation rate L is not only related with
the eddy dissipation due to viscosity similar to the single-phase turbulent flow but
also the dispersed bubbles B due to the bubble deformation, the capture of the
bubbles by eddies and also the collisions between the bubbles and eddies. The
contribution of the dispersed particles to the dissipation rate L is related with the
characteristic length and time of both the dispersed bubbles and eddies, which was
discussed in [7.22]. However, the case with the large light particles such as large
bubbles was not considered in detail because of its complexity such as the deformability,
which still needs further studies in the future. Here, as an initial step the dissipation rate
L will be considered based on the effective viscosity t in which for the turbulent
bubbly flow the bubble contribution is needed to be considered:
2
kL
t
tC
(7.4)
151
where Cμ is the constant and around 0.09 for the single-phase turbulent flow.
Substituting Eq. (7.4) into Eq. (7.3), it is obtained that
rLt
kC
t VF drag
. (7.5)
At the duct center, the eddy viscosity t induced by the bubbles is considered
by the model proposed by Sato and Sadatomi [7.23],
rBbt wDC (7.6)
Substituting Eqs. (1.4), (7.2) and (7.6) into Eq. (7.5), the turbulent kinetic energy
tk in the liquid phase could be determined as follows:
b
StokesDD
rk CC
CCWt
,2
(7.7)
where StokesDC , was determined by ,Stokes 24 / ReDC ; Cμ is chosen to be 0.09
similar to the single-phase turbulent flow; Cμb is chosen to be 0.6; the correlation
proposed by Tomiyama et al. [7.24] was used to determine drag coefficients as
5.0
687.0
1
1
4Eo
Eo
3
8,
Re
72,Re15.01
Re
24minmax
B
B
B
DC .
Fig. 7.8 shows the turbulent kinetic energy tk based on Eq. (7.7) for different
bubble diameters, from which better prediction of tk could be obtained for the
turbulent bubbly flow in the large ducts. Hence, Eq. (7.7) is suggested for the
turbulent bubbly flow in the large ducts. However, the effect of the bubble
deformation still needs further consideration in the future.
152
Fig. 7.7 (a) The axial turbulent intensities <ww> vs. the local void fraction α in the duct
center; (b) <ww>/Wl vs. the local void fraction α in the duct center.
0 0.05 0.1 0.15 0.2 0.25 0.30
0.02
0.04
0.06
0.08
<w
w>
Serizawa
Sun
Shawkat
0.2 Wr
2
Wr
2
0.721.6
0 0.05 0.1 0.15 0.2 0.25 0.30
0.02
0.04
0.06
0.08
<w
w>
/Wl
Serizawa
Sun
Shawkat
0.35
0.15
0.02
(a)
(b)
153
Fig. 7.8 Comparison the experimental data of the turbulent kinetic energy tk of the liquid
phase with the new correlation Eq. (7.7) for different bubble diameter at the duct center.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
t k
Liu
Serizawa
Sun
Shawkat
0.5 Wr
2
DB
=3mm
DB
=4mm
DB
=5mm
DB
=6mm
154
7.2.2 Bubble-induced and wall-induced turbulence interaction
In the near wall region, the turbulence in the liquid phase includes the
wall-induced and bubble-induced. To investigate the bubble-induced turbulence, the
wall-induced contribution is necessary to be extracted. However, near the wall the
wall-induced turbulence could be strongly modified by the bubbles, thus it might be not
suitable to directly extract the wall-induced single-phase turbulence from the total
turbulence of the turbulent bubbly flow. In other words, it is very difficult to directly
investigate the bubble-induced turbulence in the near wall region. In the present study,
the bubble-induced turbulence and wall-induced turbulence will be investigated near the
wall in an indirect way hereafter.
7.2.2.1 Wall effect on bubble-induced turbulence
As discussed in Section 7.2.1, the turbulence in the duct center could be
considered solely induced by bubbles. Firstly, the ratio of the local turbulent kinetic
energy tkt to that in the duct center c
ktt will be analyzed and shown in Figs. 7.9-7.13 for
the circular pipes with small and large diameter and also the noncircular ducts based on
the existed experimental databases. The following trends could be observed:
(1) From Figs. 7.9-7.13, except two cases under Jl=0.491m/s from Hibiki’s
experimental databases [7.15] in the transition regime, generally the ratio tkt /
c
ktt for
the single-phase turbulent flow is much larger than that for the turbulent bubbly flow
even with higher void fraction near the wall, i.e.
cB
kb
wB
kb
wB
kw
cB
kb
cB
kw
wB
kb
wB
kw
cS
kw
wS
kw
t
tt
tt
tt
t
t,
,,
,,
,,
,
,
. (7.8)
where wS
kwt , and
cS
kwt , are the wall-induced turbulence near the wall and in the duct
center for the single-phase turbulent flow; wB
kwt , and
cB
kwt , are the wall-induced
turbulence near the wall and in the duct center for the turbulent bubbly flow; wB
kbt , and
cB
kbt , are the bubble-induced turbulence near the wall and in the duct center for the
155
turbulent bubbly flow. The exceptions in the small circular pipes mentioned above
might be attributed to the bubble coalescence and breakup which induced another
mechanism for the turbulence generation, which will not be considered here.
Rearranging Eq. (7.7), it could be obtained that
cS
kw
cB
kb
wS
kw
wB
kw
t
t
t
t,
,
,
,
, (7.9)
indicating that for the turbulent bubbly flow the modification of the wall-induced
turbulence is much less than that in the duct center. In addition, from Eq. (7.8) we also
have
cB
kb
wB
kb
cB
kb
wB
kb
wB
kw
t
t
t
tt,
,
,
,,
. (7.10)
Based on Eq. (7.10), the wall effect on the bubble-induced turbulence could be
considered. Due to the non-uniform void fraction distribution, we need to know how
much is the bubble-induced turbulence in the absence of the wall before in order to
consider the wall effect on the bubble-induced turbulence, which will be considered by
the proposed correlation Eq. (7.8) and the linear relation with the void fraction, i.e.,
c
w
cB
kb
wB
kb
t
t
,
,
. (7.11)
Substituting Eq. (7.11) to Eq. (7.10), it should be
c
w
cB
kb
wB
kb
wB
kw
t
tt
,
,,
. (7.12)
Fig. 7.14 shows the comparison between the left and right hands of Eq. (7.11)
based on Sun’s experimental database. In contrast, it is observed that
c
w
cB
kb
wB
kb
wB
kw
cB
kb
wB
kb
t
tt
t
t
,
,,
,
,
, (7.13)
indicating that comparing to the core region under the same void fraction the
156
bubble-induced turbulence is reduced by the wall effect. The following two mechanisms
could be considered to be responsible for this observation: (a) the wall effect elongates
the bubble shape; (b) the wall suppressed the bubble wake development.
The scattering in Fig. 7.14 is due to the different liquid flow rates, i.e.,
l
c
w
cB
kb
wB
kb
wB
kw JFt
tt,
,
,,
. (7.14)
Considering / ,w c l cF J H as shown in Fig. 7.15(a), Fig. 7.15(b)
shows comparison between , , ,/B w B w B c
kw kb kbt t t and c , in which the scattering was
significantly improved. Besides, it was found that , , ,/B w B w B c
kw kb kbt t t linearly decreases
with the increase of c , indicating the turbulence near the wall could be predicted by the
following correlation:
c
cB
kb
wB
kb Htt ,, . (7.15)
cH shows the suppression of the bubble-induced turbulence and is related
with duct geometry and the local void fraction. It still needs further researches and
verification for different cases in the future.
From Figs. 7.9-7.13, it was also observed that for the single-phase turbulent flow
the ratio tkt/c
ktt is always larger than 1, since the turbulence is mainly generated near the
wall and decayed to the duct center. However, it is not always true for the turbulent
bubbly flow even for the cases with higher void fraction near the wall, i.e.,
cB
kb
wB
kb
wB
kw
cB
kb
cB
kw
wB
kb
wB
kw
cS
kw
wS
kw
t
tt
tt
tt
t
t,
,,
,,
,,
,
,
1
or w
wB
kb
wB
kwc
cB
kb ttt ,,, . (7.16)
For the turbulent bubbly flow in the large ducts as shown in Figs. 7.12-7.13
where the void fraction shows a flatter distribution, this tendency could be obviously
observed and further confirms wall suppression of the bubble-induced turbulence. In the
small circular pipe, increasing the void fraction when the near wall turbulence was
dominated by the bubbles, this phenomenon could be observed as shown in Fig. 7.11.
157
Fig. 7.9 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to that in
the duct center c
ktt
in the small circular pipe with DH=60mm based on Serizawa’s
experimental database: under (a) Jl =0.74m/s and (b) Jl =1.03m/s [7.2].
Fig. 7.10 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to that
in the duct center c
ktt
in the small circular pipe with DH=50.8mm under (a) Jl
=0.491m/s and (b) Jl =0.986m/s based on Hibiki’s experimental database [7.15].
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
r/R
t k/t
c k
Jl=0.74m/s X %
0.000
0.0119
0.024
0.0358
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
r/R
t k/t
c k
Jl=1.03m/s X %
0.000
0.0085
0.0170
0.0258
0.0341
0.0427
0.0511
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
r/R
t kt/t
c kt
Jl=0.491m/s J
g
0.000
0.0275
0.0556
0.129
0.190
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
r/R
t kt/t
c kt
Jl=0.986m/s J
g
0.000
0.0473
0.113
0.242
0.321
(a) (b)
(a) (b)
158
Fig. 7.11 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to that
in the duct center c
ktt
in the small circular pipe with DH=60mm based on Liu’s
experimental database [7.7].
Fig. 7.12 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to that
in the duct center
c
ktt in the large circular pipe with DH=200mm under (a) Jl =0.45m/s
and (b) Jl =0.68m/s based on Shakwat’s experimental database [7.13].
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
7
r/R
t kt/t
c kt
Jl J
g
0.376 0.00
0.753 0.00
0.376 0.18
0.535 0.18
0.753 0.18
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
r/R
t kt/t
c kt
Jl=0.45m/s J
g
0.0000.0150.030
0.0650.0850.100
0 0.2 0.4 0.6 0.8 10
1
2
3
4
5
6
7
8
r/R
t kt/t
c kt
Jl=0.68m/s J
g
0.0000.0150.0500.065
0.0850.1000.180
(a) (b)
159
Fig. 7.13 Lateral distributions of the ratio of the local turbulent kinetic energy tkt to that in the duct center
c
ktt in the large noncircular
pipe with DH=136mm based on Sun’s experimental database: under Jl =0.75m/s (a) along the bisector line; (b) along the diagonal line;
under Jl =1.0m/s (c) along the bisector line; (b) along the diagonal line [7.16].
0 0.2 0.4 0.6 0.8 10
2
4
6
8
10
t kt /
t kt
c
Jl=0.75m/s Bisector
Jg
2x/DH
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
2
4
6
8
10
t kt /
t kt
c
Jl=0.75m/s Diagonal
Jg
2x/DH
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
2
4
6
8
10
t kt /
t kt
c
Jl=1.0m/s Bisector
Jg
2x/DH
0.0000.0900.135
0.1800.225
0 0.2 0.4 0.6 0.8 10
2
4
6
8
10
t kt /
t kt
c
Jl=1.0m/s Diagonal
Jg
2x/DH
0.0000.0900.135
0.1800.225
(a) (b)
(c) (d)
160
Fig. 7.14 Comparison between , , ,/B w B w B c
kw kb kbt t t and /w c based on Sun’s
experimental database [7.16].
1 2 3 4 5 60
1
2
3
4
5
6
w
/c
(tk
w
B,w
+t k
b
B,w
) /t
kb
B,c
Jl=0.50
Jl=0.75
Jl=1.00
Jl=1.25
(tkw
B,w+t
kb
B,w) /t
kb
B,c=
w /
c
161
Fig. 7.15 (a) Relation between /w c and c ; (b) Relation between
, , ,/B w B w B c
kw kb kbt t t and c based on Sun’s experimental database [7.16].
0 0.05 0.1 0.15 0.21
2
3
4
5
6
w
/
c
Jl=0.50
Jl=0.75
Jl=1.00
Jl=1.25
0 0.05 0.1 0.15 0.20
2
4
6
8
10
c
(tk
w
B,w
+t k
b
B,w
) /t
kb
B,c
Jl=0.50
Jl=0.75
Jl=1.00
Jl=1.25
(a)
(b)
162
7.2.2.2 Bubble effect on wall-induced turbulence
It is difficult to consider the bubble effect on wall-induced turbulence near the
wall coupling with the modification of the bubble-induced turbulence by the wall effect.
As compared to the single-phase turbulent flow, generally the turbulence in the liquid
phase increases with the void fraction for the turbulent bubbly flow, the tendency of
which is shown in Fig. 7.16. For cases with low void fraction, the turbulence reduction
in the liquid phase was found as compared to the single-phase turbulent flow [7.8].
Fig.7.17 shows the experimental conditions under which turbulence reduction in the
liquid phase was observed. Hereafter, a simple criterion for turbulent reduction will be
considered in the turbulent bubbly flow.
For the case with the turbulence reduction in the liquid phase, we have
wS
kww
wB
kb
wB
kw ttt ,,, . (7.17)
As for the bubble-induced turbulence, though it was suppressed by the wall
effect, it should be stronger than that the pseudo-turbulence based on the potential
theory, i.e.,
2,
2
1Rww
wB
kb Wt . (7.18)
Substituting Eq. (7.18) to Eq. (7.17), we have
2,,
2
1Rw
wS
kw
wB
kw Wtt . (7.19)
Based on Eq. (7.19), the critical void fraction near the wall could be obtained as
2, /2 R
wS
kwcalc Wt . It indicates that the turbulent enhancement will occur if
2, /2 R
wS
kww Wt and the turbulent reduction is only possible to occur if 2, /2 R
wS
kww Wt .
Fig.7.18 shows comparison the experimental measured void fraction and the calculated
void fraction based on 2, /2 R
wS
kwcalc Wt near the wall for the cases including
turbulence enhancement and reduction. Since it is lack of the experimental database of
163
the void fraction and near wall turbulence, only the data from Shakwat et al. [7.13] and
Sun [7.16] were shown, based on which the above consideration was confirmed.
Moreover, it explains the reason why the turbulence reduction could occur under higher
void fraction for larger liquid flow rate but lower void fraction for smaller liquid flow
rate.
As for the mechanism for the turbulence reduction near the wall under certain
conditions, in [7.8] it was attributed to the following two effects due to the bubbles: (1)
large velocity fluctuations gradient near the bubble interface which increases the
turbulent energy dissipation; (2) energy dumping effect due to the bubble deformation.
Here, another two mechanisms responsible for the reduction of the wall-induced
turbulence in the turbulent bubbly flow are proposed as shown in Fig. 7.19.
(i) Flow acceleration laminarization
As shown in Fig. 7.19(a), due to the formation of the bubble layer near the wall,
the liquid flow was confined between the wall and the bubble layer, which laminarized
the flow due to the local flow acceleration due to the bubbles. However, it still needs
further experimental measurement including the velocity and turbulence in the inner
layer to confirm the above points.
(ii) Suppression of “bursting event” of the coherent structures
For the single-phase flow, the turbulence generation is mainly due to the
―bursting event‖ of the coherent structures near the wall as shown in shown in Fig.
7.19(b). However, the presence of the bubbles whose size is on the similar order of the
coherent structures broke these structures, thus suppressed the source of the
wall-induced turbulence.
164
Fig. 7.16 Schematic diagram of the relation between the turbulence in the liquid phase
and the void fraction for the turbulent bubbly flow.
Fig. 7.17 Experimental conditions with turbulence suppression and enhancement in the
liquid phase from [7.8].
10-3
10-2
10-1
100
10-1
100
101
Jg m/s
Jl m
/s
Serizawa [7.2,7.8]
Inoue [7.25]
Sullivan [7.26]
Ohba [7.27]
Aoki [7.28]
Lee et [7.29]
Shakwat [7.13]
Sun [7.16]
Turbulence Reduction
Turbulence Enhancement
165
Fig. 7.18 Comparison the experimental measured void fraction and the calculated void
fraction based on 2, /2 R
wS
kwcalc Wt near the wall.
Fig. 7.19 Schematic diagram of the mechanism for the turbulence suppression.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.350
0.05
0.1
0.15
0.2
0.25
0.3
0.35
calc
ex
p
Shakwat [7.13] turbulence reduction
Sun [7.16] turbulence enhancement
Shakwat [7.13] turbulence enhancement
Turbulence enhancement
occurs.
Turbulence reduction is
possible to occur.
166
7.3 Turbulence anisotropy in turbulent bubbly flow
As mentioned in Chapter 1, so far there are very few detailed discussions on the
turbulence anisotropy of the turbulent bubbly flow, based on which the qualitative
effects of the bubbles on the turbulence anisotropy are still contradictory [7.6, 7.9,
7.30-7.31]. Therefore, how the bubbles affect turbulence anisotropy and how to model
turbulence anisotropy are still two open questions. This section is aimed at these two
topics. Firstly, the bubble effects on turbulence anisotropy will be analyzed in the liquid
phase based on the currently established database for bubbly flows in circular and
noncircular ducts. According to the analysis, the suggestions on the future model of the
turbulence anisotropy will be given then.
7.3.1 Bubble effects on turbulence anisotropy in liquid phase
Firstly, the turbulence anisotropy tensor b was defined to show the strength of
turbulence anisotropy as follows:
/ 2 1 / 3 / 2 / 2
/ 2 / 2 1 / 3 / 2
/ 2 / 2 / 2 1 / 3
k k k
k k k
k k k
uu t uv t uw t
uv t vv t vw t
uw t vw t ww t
b . (7.20)
where b11+ b22+ b33=0; for the isotropic turbulence, b=0. Here, the normal
components b11, b22, and b33 will be focused on. The deviation from 0 shows the
strength of the turbulence anisotropy.
Figs. 7.20-7.24 show b11, b22, and b33 calculated based on the current database
for the circular pipes with small and large diameter and the noncircular ducts. The
following trends could be observed:
(1) For the single-phase turbulent flow, the turbulence anisotropy increase
from the duct center to the near wall region, due to the high velocity
gradient and the wall suppression on the turbulent wall-normal
167
components in the near wall region.
(2) For the turbulent bubbly flow, as shown in Fig. 7.20 and Fig. 7.23,
Serizawa’s experiments [7.2] in the smaller circular pipes and Shakwat’s
experiments [7.13] with lower gas flow rates in the large circular pipe
showed the introduction of the bubbles didn’t alter the turbulent isotropy
or made the turbulence more isotropic in the whole cross section.
However, as shown in Fig. 7.21 and 7.22, experiments of Wang et al.
[7.6] and Liu and Bankoff [7.7] in the smaller circular pipes showed that
the presence of the bubbles made the turbulence more anisotropic in the
core region but more isotropic in the near wall region. Similar results
were also obtained for the turbulent bubbly flow in the large circular pipe
based on Shakwat’s experiments [7.13] with the higher gas flow rates and
in the large non-circular duct based on Sun’s experiments [7.16].
(3) For the cases as shown in Figs. 7.21-7.24 in which bubbles increased the
turbulence anisotropy, in the core region the anisotropy tensor was found
to reach a constant value to some extent. For example 22.011 b and
11.033 b where b11 and b33 are in the axial and lateral direction,
respectively.
In order to understand the above observations and model the bubble effects on
the turbulence anisotropy in liquid phase, it is necessary to discuss the following
parameters’ effects due to their crucial importance to the bubble-induced turbulence
including: (a) void faction; (b) duct geometry; (c) bubble size; (d) bubble relative
velocity; (e) bubble deformation.
Firstly, the void fraction effect is considered in the duct center, where the
turbulence might be dominated by bubbles. Fig.7.25 shows the relation between the
turbulence anisotropy tensor b11 and b33 and the local void fraction α in the duct center
168
for different ducts. Based on the potential theory [7.11], the anisotropy tensor of the
bubble-induced turbulence is as follows and shown in Fig. 7.25 as a reference:
1/ 30 0 0
0 1/ 30 0
0 0 1/15
b . (7.21)
For the turbulent bubbly flow in the large ducts based on Shakwat’s [7.13] and
Sun’s [7.16] experiments, it could be obtained that the turbulence is more isotropic for
the small void say less than 5% and the anisotropy tensor b was closer to the value
predicted based on the potential theory in Eq. (7.20), which is in consistent with that
observed by Lance [7.9]. For the higher void fraction, the turbulence is anisotropic and
the anisotropy tensor b reaches a constant value independent of the void fraction.
For the turbulent bubbly flow in the small circular pipes, Serizawa’s data
showed more isotropic turbulence closer to the potential theory and was focused in the
smaller void fraction range say less than 10%. However, Wang’s [7.6] and Liu’s [7.7]
data in the small circular pipes showed more anisotropic turbulence similar to that in the
large ducts, which were focused in the larger void fraction range larger than 10%.
Therefore, how the bubble affects the turbulence anisotropy is related with the
void fraction, and the turbulence anisotropy increase with the void fraction to a constant
value. The tendency of the void fraction effect on the turbulence anisotropic tensor b
obtained here is inconsistent with that in Trygovasson’s DNS [7.31] results for the
solely bubble-induced turbulence. In [7.31], it showed the turbulence behaves more
isotropic with increase of the void fraction by comparing the results for 2%, 6%, and
12%, which might be due to the reduction of the relative velocity with increase of void
fraction.
In addition, how the bubble affects the turbulence anisotropy is also related with
the duct geometry. As obtained from Fig. 7.25, the critical void fraction under which the
turbulence anisotropy reached to the constant value is larger for the smaller pipes. It
could be attributed to the strong wall effect therein.
169
Next, the bubble size effect is considered in the duct center as shown in Fig. 7.26.
Since there was only few database of the turbulence together with the bubble size, the
rough range of the bubble size was shown in Fig. 7.26. For the current database, the
bubble size DB was around 3mm in the small circular pipes, and 3~6mm in the large
ducts. It was observed from the database of the large ducts, the turbulence was more
isotropic for the smaller bubble size but more anisotropic for larger bubble size. It might
also explain why the critical void fraction under which the turbulence anisotropy
reached to the constant value is larger for the smaller pipes, which might be due to the
smaller bubble size therein.
In summary, Figs. 7.25-7.26 indicate that in the core region when the local
turbulence was dominated by the bubble-induced turbulence, under high void fraction
and larger bubble size the turbulence will be more anisotropic and the anisotropy tensor
reaches a constant value to some extent with 22.011 b and 11.033 b .
As for the bubble relative velocity and the bubble deformation effects, since there
is very few data to identify these effects on the turbulence anisotropy, in the following it
is qualitatively considered. The turbulence anisotropy in the turbulent bubbly flow could
be considered to be determined by the energy production among different direction by
the bubbles. For higher relative velocity the bubble-induced turbulence will be stronger
and more energy will be injected to the axial direction, which could make the turbulence
more anisotropic. In addition, Trygovasson’s DNSs [7.31] results showed that as
compared to the spherical bubbles, the flow with deformed bubbles behaves more
anisotropic because of the capture of the deformed bubbles in the formed bubble wake.
170
Fig. 7.20 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=60mm based on Serizawa’s experimental database: under
(a) Jl =0.45m/s and (b) Jl =0.61m/s [7.2].
Fig. 7.21 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=57.2mm based on Liu’s experimental database [7.7].
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b3
3
Jl=0.45m/s
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg
0.000
0.035
0.053
0.078
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b3
3
Jl=0.61m/s
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg
0.000
0.047
0.107
0.136
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b3
3
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg
0.376 0.000
0.753 0.000
0.376 0.018
0.535 0.018
0.753 0.018
(a) (b)
171
Fig. 7.22 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the
small circular pipe with DH=57.15mm based on Wang’s experimental database [7.6].
Fig. 7.23 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the large
circular pipe with DH=200mm based on Shakwat’s experimental database: under (a) Jl
=0.45m/s and (b) Jl =0.68m/s [7.13].
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b3
3
Jl=0.43m/s
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg 0.0
0.10
0.27
0.43
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
r/R
b3
3
Jl=0.45m/s
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg
0.000
0.015
0.085
0.100
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
r/R
b3
3
Jl=0.68m/s
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
r/R
b1
1
Jg
0.000
0.015
0.050
0.100
0.180(a) (b)
172
Fig. 7.24 Lateral distributions of the turbulence anisotropy tensor b11 and b33 in the large circular pipe with DH=136mm based on Sun’s
experimental database: under Jl =0.75m/s (a) along the bisector line; (b) along the diagonal line; under Jl =1.0m/s (c) along the bisector
line; (b) along the diagonal line [7.16].
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b33
Jl=0.75m/s Bisector
0 0.2 0.4 0.6 0.8 1
-0.4
-0.3
-0.2
-0.1
02x/D
H
b11
Jg
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b33
Jl=0.75m/s Diagonal
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
Jg
2x/DH
b11
0.0000.0680.090
0.1350.180
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b33
Jl=1.0m/s Bisector
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
2x/DH
b11
Jg
0.0000.0900.135
0.1800.225
0 0.2 0.4 0.6 0.8 10
0.1
0.2
0.3
0.4
b33
Jl=1.0m/s Diagonal
0 0.2 0.4 0.6 0.8 1-0.4
-0.3
-0.2
-0.1
0
2x/DH
b11
Jg
0.0000.0900.135
0.1800.225
(a) (b)
(c) (d)
173
Fig. 7.25 The relation between the turbulence anisotropy tensor b11 and b33 and the local
void fraction α in the duct center for different ducts. The black and green symbols
indicate the results in the small circular pipes and the large ducts respectively.
Fig. 7.26 The relation between the turbulence anisotropy tensor b11 and b33 and the local
bubble size α in the duct center for different ducts. The black and green symbols
indicate the results in the small circular pipes and the large ducts respectively.
0
0.1
0.2
0.3
0.4
b3
3
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4-0.4
-0.3
-0.2
-0.1
0
b1
1
Serizawa
Wang
Liu
Shawkat
Sun
b33
=-0.11
b33
=-1/30
0
0.1
0.2
0.3
0.4
b3
3
2 3 4 5 6-0.4
-0.3
-0.2
-0.1
0
DB
mm
b1
1
Serizawa
Liu
Shakawat
Sun
b33
=-0.11
b33
=-1/30
174
7.3.2 Modeling of the anisotropic turbulent bubbly flow
In this section, we will focus on how to model the anisotropic turbulence in the
liquid phase of the turbulent bubbly flow. Before this, the physical process of the
turbulence anisotropy generation is considered for the turbulent bubbly flow as
compared to that of the single-phase turbulent flow.
7.3.2.1 Turbulence anisotropy in the single-phase turbulent flow
Firstly, transport equation of the Reynolds stress is given in Eq. (7.22) and (7.23)
ijijijij3
2DP
Dt
uuD ji (7.22)
where Pij is the production term , k
ikj
k
j
kiijx
Uuu
x
UuuP
; Dij is the diffusion
term,
k
ijk
ijx
TD
with ikjjki
k
ji
kjiijk pupux
uuuuuT
; Φij is the
redistribution term,
ijij
i
j
j
iij
x
u
x
up
3
2; εij is the dissipation rate of
jiuu , k
j
k
iij
x
u
x
u
2 and 332211 .
For the single-phase turbulent flow, the production and redistribution term plays
an important role in the turbulence anisotropy property. Models considering the slow
term (return-to-isotropy) and fast term (rapid-distortion theory) have been proposed for
the redistribution term, which is shown as follows:
ijijijkjiR PPCtuu
kC
3
2
3
22ij (7.23)
where P=P11 +P22 +P33; the first and second terms are Rotta’s model for the slow term
and isotropization of production (IP) model [7.32].
175
The turbulence anisotropy could be considered as the problem of the turbulent
energy distribution among different components. As shown in Eq. (7.21) and Fig.
7.27(a) each turbulent component is determined by the energy balance among the
production, diffusion, redistribution and dissipation. The production term due to
Reynolds shear stress and the velocity shear causes the turbulence anisotropy, while the
redistribution term due to the interaction between the pressure and strain rate of the
velocity field returns the anisotropic turbulence to isotropic and counteracts the
anisotropic production effect. Fig. 7.27(b) shows the schematic diagram of the
generation of the turbulence anisotropy of the single-phase turbulent flow in the channel.
In the near wall region, the turbulence behaves strongly anisotropic because of the wall
suppressing effect on the normal turbulent components and the existence of the high
velocity shear and Reynolds stress. While in the bulk region, the turbulence is much less
anisotropic. For the region far enough from the wall, the turbulence might be considered
as isotropic.
Fig. 7.27 (a) Contribution of each term in the energy balance to the turbulence
anisotropy; (b) Schematic diagram of the generation of the turbulence anisotropy of the
single-phase turbulent flow in channel.
176
7.3.2.2 Turbulence anisotropy for turbulent bubbly flow
As compared to the single-phase turbulent flow, the existence of the dispersed
bubbles in the turbulent bubbly flow modifies the energy balance for each turbulent
component as shown in Eq. (7.24) by introducing another mechanism of turbulence
production, diffusion, redistribution and dissipation, which affected the turbulence
anisotropy to some extent as discussed in Section 7.3.2.
2 2
3 1 3
i j S S S S B B B B
ij ij ij i j ij ij ij i j
D u uP D P D
Dt
, (7.24)
where B
ijP , B
ijD , B
ij , and B are the turbulence production, diffusion, redistribution
and dissipation terms due to bubbles.
Correspondingly, the generation mechanism of the turbulence anisotropy is
different from that in the single-phase turbulent flow. To understand the anisotropic
property of the turbulent bubbly flow and simplify the problem, the turbulence will be
considered in the area where the local turbulence is dominated by bubbles and the
shear-induced turbulence and anisotropy could be neglected such as the core region or
in the uniform flow. Therein, Eq. (7.24) could be simplified as
2 2
3 1 3
i j Slow S B B B
ij i j ij ij i j
D u uP
Dt
. (7.25)
Hereafter, the modeling of the terms related with the dispersed bubbles will be
considered.
In order to model the anisotropic turbulence property, the turbulence induced by
bubbles is considered to mimic the grid-induced turbulence as shown in Fig. 7.28,
which is often adopted to the study of the isotropic turbulence for single-phase turbulent
flow [7.32]. Therein, the turbulence was firstly induced by the wake behind the grid and
grid shear, and then return to isotropy during the turbulence decay with time. Similar to
177
that in the grid-generated turbulence, Fig. 7.29 shows the schematic diagram of
turbulence generation process with the dispersed bubbles in the core region, where the
bubbly flow was firstly simplified in the form of column like the grids. Similarly, the
turbulence is considered at the measured points as shown in Fig. 7.29. Similar to the
grids in the single-phase turbulence, there exits wake behind the bubbles and velocity
shear between the bubbles for the turbulent bubbly flow. Within the measuring time △T
in the liquid phase it might experience two periods i.e., (1) inside the bubble wake when
the turbulence Tw was produced and developed and (2) outside the bubble wake when
the turbulence decays TD and returns to isotropic. The process of the turbulence
anisotropy could be considered as the turbulence anisotropy generates in the first period
and then decays in the second period until the next bubble comes. Therefore, the final
turbulence anisotropy in the experimental measurements was strongly related with △T
and the ratio between Tw and △T, which determines the turbulence property and the
effect of return-to-isotropy. The above parameters could be considered as follows:
1cB B
B B
LlT
W W
, (7.26)
B
BB
B
wW
W
Df
W
LT
Re , (7.27)
1
ReBW f
T
T, (7.28)
where lB-B is the vertical distance between the bubbles and was obtained in Eq. 5.21; lw
is the wake length of the bubble, which is related with the bubble size DB and bubble
Reynolds number ReB.
As obtained from Eq. (7.28), Tw/△T increases with that of the void fraction. It
explains why for the turbulent bubbly flow with very low void fraction, the measured
turbulence behaves more isotropic because of the stronger effect of the
return-to-isotropy. In addition, Tw/△T increases with that of bubble size. It explains why
178
for the turbulent bubbly flow with small bubble size, the measured turbulence behaves
more isotropic because of the more isotropic turbulence generated by bubbles initially.
Therefore, during the derivation of the anisotropic turbulence model for the
turbulent bubbly flow, the above process is suggested to be considered in the turbulence
production term in the future work based on the understanding of the turbulence
anisotropy inside the bubble wake.
B B B
ij ijP A P and Re
1
B B
ij ij
fA F
. (7.29)
As for the other two terms B
ij and B , it is much more complicated and is
suggested to be considered as follows in the future model. The introduction of the
moving bubbles could be considered as the additional wall-echo effect which suppresses
the lateral turbulence and increases the turbulence anisotropy. As for the dissipation
induced by bubbles, it is suggested to be reflected in the B
ijA in the future model, i.e.,
considering the net turbulence production term due to the bubbles.
179
Fig. 7.28 Schematic diagram of grid-generated decaying homogeneous isotropic
turbulence.
Fig. 7.29 Schematic diagram of turbulence generation of the turbulent bubbly flow in
the core region where the turbulence is dominated by bubbles.
180
7.4 Summary and discussion
In this Chapter, the turbulence modulation of the turbulent bubbly flow was
studied based on the recently established experimental database including the turbulent
kinetic energy and turbulence anisotropy. The following findings were obtained:
(1) In order to quantify the solely bubble-induced turbulence in a wider range
of void fraction, the turbulence was investigated in the duct center where it
was dominated by the bubbles. It was found that the relation between the
turbulence and the void fraction was affected by the duct geometry.
Comparing to the small circular pipes the stronger turbulence could be
induced in the large ducts. In addition, the turbulence in the small circular
pipe was found to be closer to prediction of the potential theory. In the large
ducts, the turbulence model needs to be modified to get better prediction.
(2) The present study on the interaction between bubble-induced turbulence and
wall-induced turbulence indicates that comparing to the core region, the
bubble-induced turbulence under the same void fraction was reduced by the
wall effect, which made the turbulence more homogeneous in the turbulent
bubbly flow. In addition, the turbulence reduction due to existence of the
bubbles was discussed to consider the bubble-effect on wall-induced
turbulence. The acceleration laminarization and suppression of the ―bursting
events‖ of the coherent structures in the near wall region were proposed as
the mechanisms responsible for the turbulence reduction therein.
(3) The present study on the turbulence anisotropy indicates that the existence
of the bubbles could affect the turbulence anisotropy such as causing the
turbulence more anisotropic in the duct core region but more isotropic in the
near wall region. How the turbulence anisotropy is affected by bubbles is
related with the void fraction, duct geometry, and bubble size. Under higher
void fraction or larger bubble size, the turbulence anisotropy tensor was
found to be more anisotropic and reach a constant value to some extent.
181
It is notable that since the database for the liquid phase turbulence is very
limited so far, the present analysis was only carried out for several databases. Therefore,
it still needs further validation of the above observations based on new turbulence
database for the turbulent bubbly flow especially in the large ducts to consider the solely
bubble-induced turbulence in the future. In addition, for the anisotropic turbulent model,
more databases about the bubble and bubble-induced wake are also required.
182
Reference
[7.1] Hosokawa, S., Tomiyama, A., 2010, Effect of bubbles on turbulent flows in
vertical channels, 7th International Conference on Multiphase flow. FL USA.
[7.2] Serizawa, A., 1974, Fluid dynamics characteristics of two-phase flow, Doctor
Thesis, Kyoto University.
[7.3] Serizawa, A., Kataoka, I., Michiyoshi, I., 1975, Turbulence structure of air-water
bubbly flow—II. local properties, Int. J. Multiphase Flow 2, 235–246.
[7.4] Theofanous, T. G., Sullivan, J., 1982, Turbulence in two-phase dispersed flows,
J. Fluid Mechanics 116, 343-362.
[7.5] Michiyoshi, I., Serizawa, A., 1986, Turbulence in two-phase bubble flow, Nucl.
Eng. Des. 95, 253-267
[7.6] Wang, S., Lee, S., Jones, O., Lahey, Jr R. T., 1987, 3-D turbulence structure and
phase distribution measurements in bubbly two-phase flows, Int. J. Multiphase
Flow 13, 327–343.
[7.7] Liu, T.J., Bankoff, S.G., 1993, Structure of air-water bubbly flow in a vertical
pipe—I. liquid mean velocity and turbulence measurements, Int. J. Heat Mass
Transfer 36, 1049–1060.
[7.8] Serizawa, A., Kataoka, I., 1990, Turbulence suppression in bubbly two-phase
flow, Nucl. Eng. Des. 122, 1–16.
[7.9] Lance, M. , Bataille, J., 1991, Turbulence in the liquid phase of a uniform
bubbly air-water flow, J. Fluid Mech., 222, 95-118.
[7.10] Lance, M., Bataille, J., 1991, Homogeneous turbulence in bubbly flows, J.
Fluids, Eng., 113, 295-300.
[7.11] Wijngaarden, van L., 1998, On pseudo turbulence, Theoret. Comput. Fluid
Dynamics 10, 449-458.
[7.12] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2007, On the liquid turbulence
energy spectra in two-phase bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 33, 300-316.
183
[7.13] Shawakat, M. E., Ching, C. Y., Shoukri, M., 2008, Bubble and liquid turbulence
characteristics of bubbly flow in a large diameter vertical pipe, Int. J.
Multiphase Flow 34, 767-785.
[7.14] Bunner, B., Tryggvason, G., 2002, Dynamics of homogeneous bubbly flows
Part 2: velocity fluctuations, J. Fluid Mech. 466, 53-84.
[7.15] Hibiki, T., Ishii, M., Z. Xiao, 2001, Axial interfacial area transport of vertical
bubbly flows, Int. J. Heat Mass Transfer 44, 1869-1888.
[7.16] Sun, H. M., 2014, Study on upward air-water two-phase flow characteristics in a
vertical large square duct, Doctor Thesis, Kyoto University.
[7.17] Lahey Jr. R.T., 1990, The analysis of phase separation and phase distribution
phenomena using two-fluid models, Nucl. Eng. Des. 122, 17-40.
[7.18] Bertodano, Lopez M., Lahey, Jr. R.T., Jones, O. C., 1994, Phase distribution in
bubbly two-phase flow in vertical ducts, Int. J. Multiphase Flow 20, 805-818.
[7.19] Politano M. S., Carrica P. M., Converti J., 2003, A model for turbulent
polydispers two-phase flow in vertical channels, Int. J. Multiphase Flow 19,
1153-1182.
[7.20] Hosokawa, S., Tomiyama, A., 2009, Multi-fluid simulation of turbulent bubbly
pipe flows, Chem. Eng. Sci. 64, 5308-5318.
[7.21] Rzehak, R., Krepper, E., 2013, Bubble-induced turbulence: Comparison of CFD
models, Nucl. Eng. Des. 258, 57-65.
[7.22] Yokomine, T., Shimizu, A., 1995, Prediction of turbulence modulation by using
k-ε model for gas-solid flows, Advances in Multiphase Flow 191-199.
[7.23] Sato, Y., Sadatomi, M., 1981, Momentum and heat transfer in two-phase bubble
flow—I. Theory, Int. J. Multiphase Flow 7, 167-177.
[7.24] Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002, Transverse migration of
single bubbles in simple shear flows, Chem. Eng. Sci. 57, 1849-1858.
[7.25] Inoue, A., Koga, T., Yaegashi, H., 1976, Void fraction, bubble and liquid
velocity profiles in a vertical pipe, Trans. Japan Soc. Mech. Eng. 42, 2521-2531.
[7.26] Sullivan, J. P., Houze, R. N., Buenger, D. E., Theofanous, T. J., 1978,
Turbulence in two-phase flows, Proc. 2nd
CSNI Specialist Meeting, Paris,
184
583-608
[7.27] Ohba, K., Yuhara, T., 1982, Study on vertical bubble flow using laser Doppler
measurement, Trans. Japan Soc. Mech. Eng. 48, 78-85
[7.28] Aoki, S., 1982, Eddy diffusivity of momentum in bubbly flow, M.S. Thesis,
Kyoto University.
[7.29] Lee, S. G., 1986, Turbulence modeling in bubbly two-phase flow, PhD Thesis,
RPI
[7.30] Bolotnov, Igor A., 2013, Influence of bubbles on the turbulence anisotropy, J.
Fluids Eng. 135, 051301.
[7.31] Bunner, B., Tryggvason, G., 2003, Effect of bubble deformation on the
properties of bubbly flows, J. Fluid Mech. 495, 77-118.
[7.32] Pope, S. B., 2000, Turbulent flows, Cambridge University Press.
185
Chapter 8
Conclusions
8.1 Summary of dissertation
Io order to improve the understanding and modeling of the turbulent bubbly
flow especially in the large square duct, the present dissertation contains three parts.
Firstly, the state-of-the-art of the upward turbulent bubbly flow was reviewed in Chapter
1 and the existing database was collected in Chapter 2. In second part, the numerical
evaluation of the existing models was carried out based on the experimental databases
in the large square duct. The third part was contributed to improving the understanding
of the flow characteristics and the physical processes therein including the flow
developing processes, the axial velocities of the liquid phase, the bubble deformation
and the turbulence modulation in the liquid phase. The main achievements in each part
could be summarized as follows:
1) Numerical evaluation (Chap. 3)
It indicates the existed model might not be suitable for the turbulent bubbly flow
in the large noncircular ducts because of the different physical processes therein as
compared to that in the small circular pipes. The detailed analyses of the physical
process of the turbulent bubbly flow in the large noncircular ducts are suggested to
improve the model therein.
2) Flow developing process (Chap. 4)
During the developing of turbulent bubbly flow, there exist at least two stages
which are forming the bubble layer and the velocity developing based on the formed
bubble layer. During the bubble layer formation at the first stage, the length required to
186
form the bubble layer was considered and found to be less than that required to develop
the liquid phase boundary layer with the similar thickness. At the second stage, the
length required to lift up the liquid phase velocity near the wall was analyzed, and the
criterion to judge the flow developing status was suggested.
3) Axial velocity in the liquid phase (Chap. 5)
For the turbulent bubbly flow in the large square duct, the formation of the
typical M-shaped velocity profile was noticed and studied by considering the index of
flow type, the formation conditions, the peak locations of the velocity profile and
velocity characteristics. Therein, two layers with the different flow types were observed
and analyzed between the wall and the duct center, which are the pressure driven and
shear driven flows. The formation of the M-shaped velocity profile was found to be
determined the duct geometry which dominates the lateral void fraction distribution and
the wall drag resistance. Based on the flow characteristics of the turbulent bubbly flow
in the large square duct, two-layer model and analysis was suggested to improve the
understanding and modeling therein.
4) Bubble deformation (Chap. 6)
The void fraction effect on the bubble deformation for the turbulent bubbly flow
was found which showed that the bubbles behaved more prolate with the increase of the
void fraction as compared to the single-bubble case. A new correlation for the statistical
bubble deformation was proposed and validated combining the bubble size and void
fraction to consider the void fraction effect. Numerical validation of the void fraction
showed that only for the cases with better developed status, the predictions of the void
fractions were in well consistent with the experimental measurements. It indicates the
importance of knowing developing status to model the turbulent bubbly flow.
5) Turbulence modulation in the liquid phase (Chap. 7)
The quantification of the solely bubble-induced turbulence in a wider range of
void fraction was carried out by focusing the turbulence in the duct center. The relation
187
between the turbulence and the void fraction was found to be affected by the duct
geometry. As compared to the small circular pipes the stronger turbulence could be
induced in the large ducts. The interaction between bubble-induced turbulence and
wall-induced turbulence was considered by comparing the turbulence in the core region
and near wall region. It indicates that as compared to the core region, the
bubble-induced turbulence could be reduced by the wall effect. The mechanisms of the
turbulence reduction of the turbulent bubbly flow were proposed which are the flow
acceleration laminarization and suppression of the coherent structures in the near wall
region. The investigation of the turbulence anisotropy indicates that the introduction of
the bubbles could affect the turbulence anisotropy such as causing the turbulence more
anisotropic in the duct core region but more isotropic in the near wall region. However,
how the turbulence anisotropy is affected by bubbles was found to be related with the
void fraction, duct geometry, and bubble size.
8.2 Recommendation for future work
1) To understand the flow developing process
There are several interesting observations during the developing of the turbulent
bubbly flow such as two-stage developing and the almost constant thickness of the
bubble layer δB. In addition, the understanding of the developing status is also important
for the flow modeling. However, the current database is still limited to understand the
detailed developing process of the turbulent bubbly flow. Therefore, the boundary layer
flow type is suggested for the bubbly flow.
2) To understand the secondary flow
Currently, the secondary flow effect in the turbulent bubbly flow was neglected
by considering the experimental data of the lateral turbulent normal stresses. So far
there is no direct experimental information about it, which is suggested in the future
work.
188
3) To understand the bubble deformation
Currently, the void fraction effect on the bubble deformation was considered for
the mean bubble size around 3.5mm~5.5mm and limited void fraction range. For the
wider range of the bubble size and flow conditions, it still needs further study in the
future.
4) To understand the turbulence modulation
The mechanisms here proposed for the turbulence reduction of the turbulent
bubbly flow still needs further experimental measurements including the velocity and
turbulence in the near wall region to confirm.
189
Nomenclature
Latin
ai Interfacial area concentration
A Dimensionless parameter to show flow developing status
AS Anisotropy matrix related with shear-induced turbulence
AB Anisotropy matrix related with bubble-induced turbulence
b Turbulence anisotropy tensor
Cε Model constants in η-ε model of the turbulent bubbly flow
Cε1 Model constants in η-ε model of the turbulent bubbly flow
Cε2 Model constants in η-ε model of the turbulent bubbly flow
CεB Model constants in η-ε model of the turbulent bubbly flow
Cμl Model constants related with shear-induced eddy viscosity
Cμb Model constants related with bubble-induced eddy viscosity
CL Lift force coefficient
CT Turbulent dispersion force coefficient
CVM Virtual mass force coefficient
DB Bubble size
DH Duct hydraulic radius
DHB Maximum horizontal bubble dimension
D Diffusion term Reynolds stress transportation
Dmax Maximum bubble size
DSM Sauter mean diameter
DV Axial bubble dimension
E Bubble aspect ratio or bubble deformation
Eα Bubble aspect ratio for bubbly flow
190
ES Bubble aspect ratio obtained by single-bubble correlation
Eo Eötvös number based on DSM
EoH Eötvös number based on DHB
fb Bubble frequency
fl Friction factor
F Local net driving force
g Gradational acceleration vector
Jk Superficial velocity of phase-k
k Turbulent kinetic energy
kBa Asymptotic value of bubble-induced turbulent kinetic energy
le Characteristic length of the most energetic eddy
ls Mixing length of shear-induced turbulence
lb Mixing length of bubble-induced turbulence
lB-B Vertical bubble-bubble interval
lw Length of the bubble wake
LB Length from the inlet to form bubble layer δB
Lc Chord length
Lδl Length from the inlet to form liquid-phase boundary layer δl
Lp Length required
M Morton number
Mik Interfacial force exerted on phase-k
Mikd
Interfacial drag force exerted on phase-k
MikL
Lift force exerted on phase-k
Miknd
Interfacial non-drag force exerted on phase-k
MikP
Interfacial term due to pressure exerted on phase-k
MikS
Interfacial term due to shear stress exerted on phase-k
MikT
Turbulent dispersion force exerted on phase-k
191
MikV
Virtual mass force exerted on phase-k
MikW
Wall force exerted on phase-k
Mwk Wall shear force exerted on phase-k
pk Pressure field in phase-k
P Production tensor in Reynolds stress transportation
P P=trace (P)/2
Re Reynolds number
Si Interfacial tensor in turbulent kinetic energy transportation
Si Si=trace (S)/2
SB Surface area of the bubble
te Characteristic time of the eddy
tk Turbulent kinetic energy in Chapter 7
TD Time period of the bubble wake decay
Tw Time period of the bubble wake
We Weber number
vk Fluctuating velocity vector in phase-k
k kv v Reynolds stress tensor in phase-k
kV Phase averaged velocity vector in phase-k
Vk Instantaneous velocity vector in phase-k: Vk= kV + vk
VT Terminal velocity
VR Relative velocity between bubble and liquid phase
VB Volume of the bubble
Wk Axial velocity of phase-k
Wlm Maximum of Axial liquid-phase velocity
W*
Dimensionless velocity
x Coordinate component, horizontal direction
X*
Dimensionless location
192
xp Location of the liquid-phase velocity peak
y Coordinate component, horizontal direction
z Coordinate component, flow direction
Greek
αk Void fraction of phase-k
Mass quality
δl Thickness of the liquid phase boundary layer
δB Thickness of the bubble layer
ε Dissipation rate of the turbulent kinetic energy
ε Dissipation term Reynolds stress transportation
Φ Redistribution term Reynolds stress transportation
γ Ratio between maximum void fraction and that in duct center
μk Dynamic viscosity of phase-k
υb Eddy viscosity due to bubbles
υs Eddy viscosity due to shear-induced turbulence
ζ Surface tension between liquid and gas phase
ζε Model constants in k-ε model
ρk Density of phase-k
τB Bubble-induced Reynolds stress tensor in liquid phase
τk Shear stress tensor in phase-k
τkT
Turbulent Reynolds stress tensor in phase-k
Superscript
B Bubbly flow
c In the duct center
S Single-phase flow
193
w Near the wall
Subscript
0 Single-bubble cases
α Bubbly flow cases
b or B Bubble-induced
c Duct center
i Interface-induced
in Insider the bubble layer
k Phase number
g Gas phase
l or L Liquid phase
m Mean value
out Outside the bubble layer
s or S Shear-induced
t Two-phase flow
w or W Wall-induced
duct center
S Single-phase flow
w Near the wall
Subscript
0 Single-bubble cases
α Bubbly flow cases
b or B Bubble-induced
194
c Duct center
i Interface-induced
in Insider the bubble layer
k Phase number
g Gas phase
l or L Liquid phase
m Mean value
out Outside the bubble layer
s or S Shear-induced
t Two-phase flow
w or W Wall-induced